Data Visualization

Exploring and Explaining with Data

Jeffrey D. Camm James J. Cochran

Wake Forest University University of Alabama

Michael J. Fry Jeffrey W. Ohlmann

University of Cincinnati University of Iowa

Australia ● Brazil ● Canada ● Mexico ● Singapore ● United Kingdom ● United States

 This is an electronic version of the print textbook. Due to electronic rights restrictions,
some third party content may be suppressed. Editorial review has deemed that any suppressed
content does not materially affect the overall learning experience. The publisher reserves the right
to remove content from this title at any time if subsequent rights restrictions require it. For
valuable information on pricing, previous editions, changes to current editions, and alternate
formats, please visit www.cengage.com/highered to search by ISBN#, author, title, or keyword for
materials in your areas of interest.

Important Notice: Media content referenced within the product description or the product
text may not be available in the eBook version.
 Data Visualization: Exploring and © 2022 Cengage Learning, Inc.
Explaining with Data, WCN: 02-300
First Edition
Unless otherwise noted, all content is © Cengage.
Jeffrey D. Camm, James J. Cochran,
Michael J. Fry, Jeffrey W. Ohlmann
ALL RIGHTS RESERVED. No part of this work covered by the copyright
herein may be reproduced or distributed in any form or by any means,
SVP, Higher Education & Skills Product: except as permitted by U.S. copyright law, without the prior written
Erin Joyner permission of the copyright owner.
VP, Higher Education & Skills Product:
Michael Schenk For product information and technology assistance, contact us at
Product Director: Joe Sabatino Cengage Customer & Sales Support, 1-800-354-9706 or
support.cengage.com.
Senior Product Manager: Aaron Arnsparger
For permission to use material from this text or product,
Senior Learning Designer: Brandon Foltz
submit all requests online at
Senior Content Manager: Conor Allen www.cengage.com/permissions.
Digital Delivery Lead: Mark Hopkinson

Marketing Director: Danae April Library of Congress Control Number: 2021930729

Executive Marketing Manager:

Nate Anderson ISBN: 978-0-357-63134-8

IP Analyst: Ashley Maynard

Cengage
IP Project Manager: Kelli Besse 200 Pier 4 Boulevard
Production Service: MPS Limited Boston, MA 02210
USA
Designer: Chris Doughman

Cover Image Source: Cengage is a leading provider of customized learning solutions with
iStockPhoto.com/mpilecky employees residing in nearly 40 different countries and sales in more
than 125 countries around the world. Find your local representative at
www.cengage.com.

To learn more about Cengage platforms and services, register or access

your online learning solution, or purchase materials for your course, visit
www.cengage.com.

Printed in the United States of America

Print Number: 01 Print Year: 2021
 Brief Contents
ABOUT THE AUTHORS xi
PREFACE xiii

Chapter 1 Introduction 2
Chapter 2 Selecting a Chart Type 26
Chapter 3 Data Visualization and Design 76
Chapter 4 Purposeful Use of Color 128
Chapter 5 Visualizing Variability 174
Chapter 6 Exploring Data Visually 226
Chapter 7 Explaining Visually to Influence with Data 284
Chapter 8 Data Dashboards 322
Chapter 9 Telling the Truth with Data Visualization 360

References 397
Index 399
 Contents
ABOUT THE AUTHORS xi
PREFACE xiii

Chapter 1 Introduction 2
1.1 Analytics 3
1.2 Why Visualize Data? 4
Data Visualization for Exploration 4
Data Visualization for Explanation 7
1.3 Types of Data 8
Quantitative and Categorical Data 8
Cross-Sectional and Time Series Data 9
Big Data 10
1.4 Data Visualization in Practice 11
Accounting 11
Finance 12
Human Resource Management 13
Marketing 14
Operations 14
Engineering 16
Sciences 16
Sports 17
Summary 18
Glossary 19
Problems 20

Chapter 2 Selecting a Chart Type 26

2.1 Defining the Goal of Your Data Visualization 28
Selecting an Appropriate Chart 28
2.2 Creating and Editing Charts in Excel 29
Creating a Chart in Excel 30
Editing a Chart in Excel 30
2.3 Scatter Charts and Bubble Charts 32
Scatter Charts 32
Bubble Charts 33
2.4 Line Charts, Column Charts, and Bar Charts 35
Line Charts 35
Column Charts 39
Bar Charts 41
2.5 Maps 42
Geographic Maps 42
Heat Maps 44
Treemaps 45
vi Contents

2.6 When to Use Tables 47

Tables versus Charts 47
2.7 Other Specialized Charts 49
Waterfall Charts 49
Stock Charts 51
Funnel Charts 52
2.8 A Summary Guide to Chart Selection 54
Guidelines for Selecting a Chart 54
Some Charts to Avoid 55
Excel’s Recommended Charts Tool 57
Summary 59
Glossary 60
Problems 61

Chapter 3 Data Visualization and Design 76

3.1 Preattentive Attributes 78
Color 81
Form 81
Length and Width 84
Spatial Positioning 87
Movement 87
3.2 Gestalt Principles 88
Similarity 88
Proximity 88
Enclosure 89
Connection 89
3.3 Data-Ink Ratio 91
3.4 Other Data Visualization Design Issues 98
Minimizing Eye Travel 98
Choosing a Font for Text 100
3.5 Common Mistakes in Data Visualization Design 102
Wrong Type of Visualization 102
Trying to Display Too Much Information 104
Using Excel Default Settings for Charts 106
Too Many Attributes 108
Unnecessary Use of 3D 109
Summary 111
Glossary 111
Problems 112

Chapter 4 Purposeful Use of Color 128

4.1 Color and Perception 130
Attributes of Color: Hue, Saturation, and Luminance 130
Contents vii

Color Psychology and Color Symbolism 132

Perceived Color 132
4.2 Color Schemes and Types of Data 135
Categorical Color Schemes 135
Sequential Color Schemes 137
Diverging Color Schemes 139
4.3 Custom Color Using the HSL Color System 141
4.4  Common Mistakes in the Use of Color in Data
Visualization 146
Unnecessary Color 146
Excessive Color 148
Insufficient Contrast 151
Inconsistency Across Related Charts 153
Neglecting Colorblindness 153
Not Considering the Mode of Delivery 156
Summary 156
Glossary 157
Problems 157

Chapter 5 Visualizing Variability 174

5.1 Creating Distributions from Data 176
Frequency Distributions for Categorical Data 176
Relative Frequency and Percent Frequency 179
Visualizing Distributions of Quantitative Data 181
5.2  Statistical Analysis of Distributions of Quantitative
Variables 193
Measures of Location 193
Measures of Variability 194
Box and Whisker Charts 197
5.3 Uncertainty in Sample Statistics 200
Displaying a Confidence Interval on a Mean 201
Displaying a Confidence Interval on a Proportion 203
5.4 Uncertainty in Predictive Models 205
Illustrating Prediction Intervals for a Simple Linear
Regression Model 205
Illustrating Prediction Intervals for a Time Series Model 208
Summary 211
Glossary 211
Problems 213

Chapter 6 Exploring Data Visually 226

6.1 Introduction to Exploratory Data Analysis 228
Espléndido Jugo y Batido, Inc. Example 229
Organizing Data to Facilitate Exploration 230
viii Contents

6.2 Analyzing Variables One at a Time 234

Exploring a Categorical Variable 234
Exploring a Quantitative Variable 237
6.3 Relationships between Variables 242
Crosstabulation 242
Association between Two Quantitative Variables 247
6.4 Analysis of Missing Data 256
Types of Missing Data 256
Exploring Patterns Associated with Missing Data 258
6.5 Visualizing Time-Series Data 260
Viewing Data at Different Temporal Frequencies 260
Highlighting Patterns in Time Series Data 262
Rearranging Data for Visualization 266
6.6 Visualizing Geospatial Data 269
Choropleth Maps 269
Cartograms 272
Summary 273
Glossary 274
Problems 275

Chapter 7 Explaining Visually to Influence with Data 284

7.1 Know Your Audience 287
Audience Member Needs 287
Audience Member Analytical Comfort Levels 289
7.2 Know Your Message 292
What Helps the Decision Maker? 293
Empathizing with Data 294
7.3 Storytelling with Charts 300
Choosing the Correct Chart to Tell Your Story 300
Using Preattentive Attributes to Tell Your Story 304
7.4  Bringing It All Together: Storytelling
and Presentation Design 306
Aristotle’s Rhetorical Triangle 307
Freytag’s Pyramid 308
Storyboarding 311
Summary 313
Glossary 313
Problems 314

Chapter 8 Data Dashboards 322

8.1 What Is a Data Dashboard? 324
Principles of Effective Data Dashboards 325
Applications of Data Dashboards 325
Contents ix

8.2 Data Dashboards Taxonomies 327

Data Updates 327
User Interaction 327
Organizational Function 328
8.3 Data Dashboard Design 328
Understanding the Purpose of the Data Dashboard 329
Considering the Needs of the Data Dashboard’s Users 329
Data Dashboard Engineering 330
8.4 Using Excel Tools to Build a Data Dashboard 331
Espléndido Jugo y Batido, Inc. 331
Using PivotTables, PivotCharts, and Slicers to Build
a Data Dashboard 332
Linking Slicers to Multiple PivotTables 343
Protecting a Data Dashboard 346
Final Review of a Data Dashboard 347
8.5  Common Mistakes in Data Dashboard Design 348
Summary 349
Glossary 349
Problems 350

Chapter 9 Telling the Truth with Data Visualization 360

9.1 Missing Data and Data Errors 363
Identifying Missing Data 363
Identifying Data Errors 366
9.2 Biased Data 369
Selection Bias 369
Survivor Bias 372
9.3 Adjusting for Inflation 374
9.4 Deceptive Design 377
Design of Chart Axes 377
Dual-Axis Charts 381
Data Selection and Temporal Frequency 382
Issues Related to Geographic Maps 386
Summary 388
Glossary 389
Problems 389

References  397

Index 399
 About the Authors
Jeffrey D. Camm is Inmar Presidential Chair and Senior Associate Dean of Business
Analytics in the School of Business at Wake Forest University. Born in Cincinnati, Ohio,
he holds a B.S. from Xavier University (Ohio) and a Ph.D. from Clemson University. Prior
to joining the faculty at Wake Forest, he was on the faculty of the University of Cincinnati.
He has also been a visiting scholar at Stanford University and a visiting professor of business
administration at the Tuck School of Business at Dartmouth College.
Dr. Camm has published more than 45 papers in the general area of optimization applied
to problems in operations management and marketing. He has published his research in
Science, Management Science, Operations Research, INFORMS Journal on Applied
Analytics, and other professional journals. Dr. Camm was named the Dornoff Fellow of
Teaching Excellence at the University of Cincinnati, and he was the 2006 recipient of the
INFORMS Prize for the Teaching of Operations Research Practice. A firm believer in prac-
ticing what he preaches, he has served as an operations research consultant to numerous
companies and government agencies. From 2005 to 2010 he served as editor-in-chief of the
INFORMS Journal on Applied Analytics (formerly Interfaces). In 2016, Professor Camm
received the George E. Kimball Medal for service to the operations research profession, and
in 2017 he was named an INFORMS Fellow.

James J. Cochran is Associate Dean for Research, Professor of Applied Statistics, and
the Rogers-Spivey Faculty Fellow at The University of Alabama. Born in Dayton, Ohio, he
earned his B.S., M.S., and M.B.A. from Wright State University and his Ph.D. from the Uni-
versity of Cincinnati. He has been at The University of Alabama since 2014 and has been a
visiting scholar at Stanford University, Universidad de Talca, the University of South Africa,
and Pole Universitaire Leonard de Vinci.
Dr. Cochran has published more than 50 papers in the development and application of
operations research and statistical methods. He has published in several journals, including
Management Science, The American Statistician, Communications in Statistics—Theory and
Methods, Annals of Operations Research, European Journal of Operational Research, Jour-
nal of Combinatorial Optimization, INFORMS Journal on Applied Analytics, and Statistics
and Probability Letters. He received the 2008 INFORMS Prize for the Teaching of Opera-
tions Research Practice, 2010 Mu Sigma Rho Statistical Education Award, and 2016 Waller
Distinguished Teaching Career Award from the American Statistical Association. Dr. Cochran
was elected to the International Statistics Institute in 2005, named a Fellow of the American
Statistical Association in 2011, and named a Fellow of INFORMS in 2017. He also received
the Founders Award in 2014 and the Karl E. Peace Award in 2015 from the American Statis-
tical Association, and he received the INFORMS President’s Award in 2019.
A strong advocate for effective operations research and statistics education as a means
of improving the quality of applications to real problems, Dr. Cochran has chaired teaching
effectiveness workshops around the globe. He has served as an operations research consul-
tant to numerous companies and not-for-profit organizations. He served as editor-in-chief of
INFORMS Transactions on Education and is on the editorial board of INFORMS Journal on
Applied Analytics, International Transactions in Operational Research, and Significance.

Michael J. Fry is Professor of Operations, Business Analytics, and Information Systems

(OBAIS) and Academic Director of the Center for Business Analytics in the Carl H. Lindner
College of Business at the University of Cincinnati. Born in Killeen, Texas, he earned a B.S.
from Texas A&M University and M.S.E. and Ph.D. degrees from the University of Michigan.
He has been at the University of Cincinnati since 2002, where he served as Department Head
from 2014 to 2018 and has been named a Lindner Research Fellow. He has also been a visit-
ing professor at Cornell University and at the University of British Columbia.
xii About the Authors

Professor Fry has published more than 25 research papers in journals such as Opera-
tions Research, Manufacturing and Service Operations Management, Transportation Sci-
ence, Naval Research Logistics, IIE Transactions, Critical Care Medicine, and Interfaces.
He serves on editorial boards for journals such as Production and Operations Management,
INFORMS Journal on Applied Analytics (formerly Interfaces), and Journal of Quantitative
Analysis in Sports. His research interests are in applying analytics to the areas of supply chain
management, sports, and public-policy operations. He has worked with many different orga-
nizations for his research, including Dell, Inc., Starbucks Coffee Company, Great American
Insurance Group, the Cincinnati Fire Department, the State of Ohio Election Commission, the
Cincinnati Bengals, and the Cincinnati Zoo and Botanical Gardens. In 2008, he was named a
finalist for the Daniel H. Wagner Prize for Excellence in Operations Research Practice, and
he has been recognized for both his research and teaching excellence at the University of
Cincinnati. In 2019, he led the team that was awarded the INFORMS UPS George D. Smith
Prize on behalf of the OBAIS Department at the University of Cincinnati.

Jeffrey W. Ohlmann is Associate Professor of Business Analytics and Huneke Research

Fellow in the Tippie College of Business at the University of Iowa. Born in Valentine,
Nebraska, he earned a B.S. from the University of Nebraska and M.S. and Ph.D. degrees
from the University of Michigan. He has been at the University of Iowa since 2003.
Professor Ohlmann’s research on the modeling and solution of decision-making prob-
lems has produced more than two dozen research papers in journals such as Operations
Research, Mathematics of Operations Research, INFORMS Journal on Computing, Trans-
portation Science, and European Journal of Operational Research. He has collaborated with
organizations such as Transfreight, LeanCor, Cargill, the Hamilton County Board of Elec-
tions, and three National Football League franchises. Because of the relevance of his work
to industry, he was bestowed the George B. Dantzig Dissertation Award and was recognized
as a finalist for the Daniel H. Wagner Prize for Excellence in Operations Research Practice.
 Preface
D ata Visualization: Exploring and Explaining with Data is designed to introduce best
practices in data visualization to undergraduate and graduate students. This is one
of the first books on data visualization designed for college courses. The book contains
material on effective design, choice of chart type, effective use of color, how to explore
data visually, how to build data dashboards, and how to explain concepts and results
visually in a compelling way with data. In an increasingly data-driven economy, these
concepts are becoming more important for analysts, natural scientists, social scientists,
engineers, medical professionals, business professionals, and virtually everyone who
needs to interact with data. Indeed, the skills developed in this book will be helpful to
all who want to influence with data or be accurately informed by data.
The book is designed for a semester-long course at either the undergraduate or graduate
level. The examples used in this book are drawn from a variety of functional areas in the
business world including accounting, finance, operations, and human resources as well as
from sports, politics, science, medicine, and economics. The intention is that this book will
be relevant to students at either the undergraduate or graduate level in a business school as
well as to students studying in other academic areas.
Data Visualization: Exploring and Explaining with Data is written in a style that does
not require advanced knowledge of mathematics or statistics. The first five chapters cover
foundational issues important to constructing good charts. Chapter 1 introduces data visual-
ization and how it fits into the broader area of analytics. A brief history of data visualization
is provided as well as a discussion of the different types of data and examples of a variety of
charts. Chapter 2 provides guidance on selecting an appropriate type of chart based on the
goals of the visualization and the type of data to be visualized. Best practices in chart design,
including discussions of preattentive attributes, Gestalt principles, and the data-ink ratio, are
covered in Chapter 3. Chapter 4 discusses the attributes of color, how to use color effectively,
and some common mistakes in the use of color in data visualization. Chapter 5 covers the im-
portant topic of visualizing and describing variability that occurs in observed values. Chapter
5 introduces the visualization of frequency distributions for categorical and quantitative vari-
ables, measures of location and variability, and confidence intervals and prediction intervals.
Chapters 6 and 7 cover how to explore and explain with data visualization in detail with
examples. Chapter 6 discusses the use of visualization in exploratory data analysis. The ex-
ploration of individual variables as well as the relationship between pairs of variables is con-
sidered. The organization of data to facilitate exploration is discussed as well as the effect of
missing data. The special considerations of visualizing time series data and geospatial data
are also presented. Chapter 7 provides important coverage of how to explain and influence
with data visualization, including knowing your message, understanding the needs of your
audience, and using preattentive attributes to better convey your message. Chapter 8 is a
discussion of how to design and construct data dashboards, collections of data visualizations
used for decision making. Finally, Chapter 9 covers the responsible use of data visualization
to avoid confusing or misleading your audience. Chapter 9 addresses the importance of
understanding your data in order to best convey insights accurately and also discusses how
design choices in a data visualization affect the insights conveyed to the audience.
This textbook can be used by students who have previously taken a basic statistics course
as well as by students who have not had a prior course in statistics. The two most techni-
cal chapters, Chapters 5 (Visualizing Variability) and 6 (Exploring Data Visually), do not
assume a previous course in statistics. All technical concepts are gently introduced. For
students who have had a previous statistics class, the statistical coverage in these chapters
provides a good review within a treatment where the focus is on visualization. The book of-
fers complete coverage for a full course in data visualization, but it can also support a basic
statistics or analytics course. The following table gives our recommendations for chapters to
use to support a variety of courses.
xiv Preface

Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter Chapter

1 2 3 4 5 6 7 8 9

Intro Chart Type Design Color Variability Exploring Explaining Dashboards Truth

Full Data Visualiza-

tion Course • • • • • • • • •
Data Visualization
Course Focused on • • • • • •
Presentation

Part of a Basic Statis-

tics Course • • • • • •
Part of an Analytics
Course • • • • • •

Features and Pedagogy

The style and format of this textbook are similar to our other textbooks. Some of the specific
features that we use in this textbook are listed here.
●● Data Visualization Makeover: With the exception of Chapter 1, each chapter contains a
Data Visualization Makeover. Each of these vignettes presents a real visualization that
can be improved using the principles discussed in the chapter. We present the original
data visualization and then discuss how it can be improved. The examples are drawn
from many different organizations in a variety of areas including government, retail,
sports, science, politics, and entertainment.
●● Learning Objectives: Each chapter has a list of learning objectives of that chapter. The
list provides details of what students should be able to do and understand once they
have completed the chapter.
●● Software: Because of its widespread use and ease of availability, we have chosen
Microsoft Excel as the software to illustrate the best practices and principles contained
herein. Excel has been thoroughly integrated throughout this textbook. Whenever we
introduce a new type of chart or table, we provide detailed step-by-step instructions
for how to create the chart or table in Excel. Step-by-step instructions for creating
many of the charts and tables from the textbook using Tableau and Power BI are also
available in MindTap.
●● Notes and Comments: At the end of many sections, we provide Notes and Comments
to give the student additional insights about the material presented in that section.
Additionally, margin notes are used throughout the textbook to provide insights and
tips related to the specific material being discussed.
●● End-of-Chapter Problems: Each chapter contains at least 15 problems to help the stu-
dent master the material presented in that chapter. The problems are separated into
Conceptual and Applications problems. Conceptual problems test the student’s under-
standing of concepts presented in the chapter. Applications problems are hands-on and
require the student to construct or edit charts or tables.
●● DATAfiles and CHARTfiles: All data sets used as examples and in end-of-chapter
problems are Excel files designated as DATAfiles and are available for download by
the student. The names of the DATAfiles are called out in margin notes throughout the
textbook. Similarly, some Excel files with completed charts are available for download
and are designated as CHARTfiles.
Preface xv

MindTap
MindTap is a customizable digital course solution that includes an interactive eBook,
auto-graded exercises and problems from the textbook with solutions feedback, interactive
visualization applets with quizzes, chapter overview and problem walk-through videos, and
more! MindTap also includes step-by-step instructions for creating charts and tables from
the textbook in Tableau and Power BI. Contact your Cengage account executive for more
information about MindTap.

Instructor and Student Resources

Additional instructor and student resources for this product are available online. Instructor
assets include an Instructor’s Manual, Educator’s Guide, PowerPoint® slides, a Solutions
and Answers Guide, and a test bank powered by Cognero®. Student assets include data sets.
Sign up or sign in at www.cengage.com to search for and access this product and its online
resources.

ACKNOWLEDGMENTS
We would like to acknowledge the work of reviewers who have provided comments and
suggestions for improvement of this first edition of this text. Thanks to:
Xiaohui Chang
Oregon State University
Wei Chen
York College of Pennsylvania
Anjee Gorkhali
Susquehanna University
Rita Kumar
Cal Poly Pomona
Barin Nag
Towson University
Andy Olstad
Oregon State University
Vivek Patil
Gonzaga University
Nolan Taylor
Indiana University

We are also indebted to the entire team at Cengage who worked on this title: Senior Prod-
uct Manager, Aaron Arnsparger; Senior Content Manager, Conor Allen; Senior Learning
Designer, Brandon Foltz; Digital Delivery Lead, Mark Hopkinson; Associate Subject-Matter
Expert, Nancy Marchant; Content Program Manager, Jessica Galloway; Content Quality
Assurance Engineer, Douglas Marks; and our Senior Project Manager at MPS Limited,
Anubhav Kaushal, for their editorial counsel and support during the preparation of this text.
The following Technical Content Developers worked on the MindTap content for this
text: Anthony Bacon, Philip Bozarth, Sam Gallagher, Anna Geyer, Matthew Holmes, and
Christopher Kurt. Our thanks to them as well.

Jeffrey D. Camm
James J. Cochran
Michael J. Fry
Jeffrey W. Ohlmann
Chapter 1
Introduction
Contents

1-1 ANALYTICS 1-4 DATA VISUALIZATION IN PRACTICE

Accounting
1-2 WHY VISUALIZE DATA? Finance
Data Visualization for Exploration Human Resource Management
Data Visualization for Explanation Marketing
Operations
Engineering
1-3 TYPES OF DATA Sciences
Quantitative and Categorical Data Sports
Cross-Sectional and Time Series Data
Big Data SUMMARY
GLOSSARY
PROBLEMS

LE A R NI N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 D
 efine analytics and describe the different types LO 3 D
 escribe various examples of data visualization
of analytics used in practice

LO 2 D
 escribe the different types of data and give LO 4 Identify the various charts defined in this chapter
an example of each
1-1 Analytics 3

You need a ride to a concert, so you select the Uber app on your phone. You enter the loca-
tion of the concert. Your phone automatically knows your location and the app presents
several options with prices. You select an option and confirm with your driver. You receive
the driver’s name, license plate number, make and model of vehicle, and a photograph of
the driver and the car. A map showing the location of the driver and the time remaining
until arrival is updated in real time.
Without even thinking about it, we continually use data to make decisions in our lives.
How the data are displayed to us has a direct impact on how much effort we must expend
to utilize the data. In the case of Uber, we enter data (our destination) and we are presented
with data (prices) that allow us to make an informed decision. We see the result of our
decision with an indication of the driver’s name, make and model of vehicle, and license
plate number that makes us feel more secure. Rather than simply displaying the time until
arrival, seeing the progress of the car on a map gives us some indication of the driver’s
route. Watching the driver’s progress on the app removes some uncertainty and to some
extent can divert our attention from how long we have been waiting. What data are pre-
sented and how they are presented has an impact on our ability to understand the situation
and make more-informed decisions.
A weather map, an airplane seating chart, the dashboard of your car, a chart of the per-
formance of the Dow Jones Industrial Average, your fitness tracker—all of these involve
the visual display of data. Data visualization is the graphical representation of data and
information using displays such as charts, graphs, and maps. Our ability to process infor-
mation visually is strong. For example, numerical data that have been displayed in a chart,
graph, or map allow us to more easily see relationships between variables in our data set.
Trends, patterns, and the distributions of data are more easily comprehended when data are
displayed visually.
This book is about how to effectively display data to both discover and describe the
information it contains data. We provide best practices in the design of visual displays of
data, the effective use of color, and chart type selection. The goal of this book is to instruct
you how to create effective data visualizations. Through the use of examples (using real
data when possible), this book presents visualization principles and guidelines for gaining
insight from data and conveying an impactful message to the audience.
With the increased use of analytics in business, industry, science, engineering, and
government, data visualization has increased dramatically in importance. We begin with a
discussion of analytics and data visualization’s role in this rapidly growing field.

1-1 Analytics
Analytics is the scientific process of transforming data into insights for making better
decisions.1 Three developments have spurred the explosive growth in the use of analytics
for improving decision making in all facets of our lives, including business, sports, science,
medicine, and government:
●● Incredible amounts of data are produced by technological advances such as point-

of-sale scanner technology; e-commerce and social networks; sensors on all kinds
of mechanical devices such as aircraft engines, automobiles, thermometers, and
farm machinery enabled by the so-called Internet of Things; and personal electronic
devices such as cell phones. Businesses naturally want to use these data to improve
the efficiency and profitability of their operations, better understand their customers,
and price their products more effectively and competitively. Scientists and engineers
use these data to invent new products, improve existing products, and make new
basic discoveries about nature and human behavior.

1
We adopt the definition of analytics developed by the Institute for Operations Research and the Management
Sciences (INFORMS).
4 Chapter 1 Introduction

●● Ongoing research has resulted in numerous methodological developments, including

advances in computational approaches to effectively handle and explore massive
amounts of data as well as faster algorithms for data visualization, machine learning,
optimization, and simulation.
●● The explosion in computing power and storage capability through better computing

hardware, parallel computing, and cloud computing (the remote use of hardware and
software over the internet) enable us to solve larger decision problems more quickly
and more accurately than ever before.
In summary, the availability of massive amounts of data, improvements in analytical meth-
ods, and substantial increases in computing power and storage have enabled the explosive
growth in analytics, data science, and artificial intelligence.
Analytics can involve techniques as simple as reports or as complex as large-scale opti-
mizations and simulations. Analytics is generally grouped into three broad categories of
methods: descriptive, predictive, and prescriptive analytics.
Descriptive analytics is the set of analytical tools that describe what has happened.
This includes techniques such as data queries (requests for information with certain charac-
teristics from a database), reports, descriptive or summary statistics, and data visualization.
Descriptive data mining techniques such as cluster analysis (grouping data points with
similar characteristics) also fall into this category. In general, these techniques summarize
existing data or the output from predictive or prescriptive analyses.
Predictive analytics consists of techniques that use mathematical models constructed
from past data to predict future events or better understand the relationships between vari-
ables. Techniques in this category include regression analysis, time series forecasting,
computer simulation, and predictive data mining. As an example of a predictive model, past
weather data are used to build mathematical models that forecast future weather. Likewise,
past sales data can be used to predict future sales for seasonal products such as snowblow-
ers, winter coats, and bathing suits.
Prescriptive analytics are mathematical or logical models that suggest a decision
or course of action. This category includes mathematical optimization models, decision
analysis, and heuristic or rule-based systems. For example, solutions to supply network
optimization models provide insights into the quantities of a company’s various products
that should be manufactured at each plant, how much should be shipped to each of the
company’s distribution centers, and which distribution center should serve each customer
to minimize cost and meet service constraints.
Data visualization is mission-critical to the success of all three types of analytics. We
discuss this in more detail with examples in the next section.

1-2 Why Visualize Data?

We create data visualizations for two reasons: exploring data and communicating/explaining a
message. Let us discuss these uses of data visualization in more detail, examine the differences
in the two uses, and consider how they relate to the types of analytics previously described.

Data Visualization for Exploration

Data visualization is a powerful tool for exploring data to more easily identify patterns,
recognize anomalies or irregularities in the data, and better understand the relationships
between variables. Our ability to spot these types of characteristics of data is much stronger
and quicker when we look at a visual display of the data rather than a simple listing.
As an example of data visualization for exploration, let us consider the zoo attendance
In chapter 2, we introduce a data shown in Table 1.1 and Figure 1.1. These data on monthly attendance to a zoo can be
variety of different chart types found in the file Zoo. Comparing Table 1.1 and Figure 1.1, observe that the pattern in the data
and how to construct charts
in Excel.
is more detectable in the column chart of Figure 1.1 than in a table of numbers. A column
chart shows numerical data by the height of the column for a variety of categories or time
periods. In the case of Figure 1.1, the time periods are the different months of the year.
 1-2 Why Visualize Data? 5

TABLE 1.1 Zoo Attendance Data

Month Jan Feb Mar Apr May Jun
Attendance 5422 4878 6586 6943 7876 17843

Month July Aug Sept Oct Nov Dec

Zoo
Attendance 21967 14542 8751 6454 5677 11422

FIGURE 1.1 A Column Chart of Zoo Attendance by Month

Attendance
25000

20000

15000

10000

5000

0
Jan Feb Mar Apr May Jun July Aug Sept Oct Nov Dec
Month

Our intuition and experience tells us that we would expect zoo attendance to be high-
est in the summer months when many school-aged children are out of school for summer
break. Figure 1.1 confirms this, as the attendance at the zoo is highest in the summer
months of June, July, and August. Furthermore, we see that attendance increases gradually
each month from February through May as the average temperature increases, and atten-
dance gradually decreases each month from September through November as the average
temperature decreases. But why does the zoo attendance in December and January not fol-
low these patterns? It turns out that the zoo has an event known as the “Festival of Lights”
that runs from the end of November through early January. Children are out of school
during the last half of December and early January for the holiday season, and this leads to
increased attendance in the evenings at the zoo despite the colder winter temperatures.
Visual data exploration is an important part of descriptive analytics. Data visualization
can also be used directly to monitor key performance metrics, that is, measure how an
Data dashboards are organization is performing relative to its goals. A data dashboard is a data visualization
discussed in more detail in tool that gives multiple outputs and may update in real time. Just as the dashboard in your
Chapter 8.
car measures the speed, engine temperature, and other important performance data as you
drive, corporate data dashboards measure performance metrics such as sales, inventory
levels, and service levels relative to the goals set by the company. These data dashboards
alert management when performances deviate from goals so that corrective actions can
be taken.
Visual data exploration is also critical for ensuring that model assumptions hold in predictive
and prescriptive analytics. Understanding the data before using that data in modeling builds
trust and can be important in determining and explaining which type of model is appropriate.
6 Chapter 1 Introduction

As an example of the importance of exploring data visually before modeling, we con-

sider two data sets provided by statistician Francis Anscombe.2 Table 1.2 contains these
two data sets, each of which contains 11 X-Y pairs of data. Notice in Table 1.2 that both
data sets have the same average values for X and Y, and both sets of X and Y also have the
same standard deviations. Based on these commonly used summary statistics, these two
data sets are indistinguishable.
Figure 1.2 shows the two data sets visually as scatter charts. A scatter chart is a
graphical presentation of the relationship between two quantitative variables. One variable
is shown on the horizontal axis and the other is shown on the vertical axis. Scatter charts
are used to better understand the relationship between the two variables under consider-
ation. Even though the two different data sets have the same average values and standard
deviations of X and Y, the respective relationships between X and Y are different.
A scatter chart is often One of the most commonly used predictive models is linear regression, which involves
referred to as a scatter plot. finding the best-fitting line to the data. In the graphs in Figure 1.2, we show the best-
fitting lines for each data set. Notice that the lines are the same for each data set. In
fact, the measure of how well the line fits the data (expressed by a statistic labeled R2)
is the same (67% of the variation in the data is explained by the line). Yet, as we can see
because we have graphed the data, in Figure 1.2a, fitting a straight line looks appropriate
for the data set. However, as shown in Figure 1.2b, a line is not appropriate for data set 2.
We will need to find a different, more appropriate mathematical equation for data set 2.
The line shown in Figure 1.2 for data set 2 would likely dramatically overestimate values
of Y for values of X less than 5 or greater than 14.
Hence, before applying predictive and prescriptive analytics, it is always best to visually
explore the data to be used. This helps the analyst avoid misapplying more complex tech-
niques and reduces the risk of poor results.

TABLE 1.2 Two Data Sets from Anscombe

Data Set 1 Data Set 2
X Y X Y
10 8.04 10 9.14
8 6.95 8 8.14
13 7.58 13 8.74
9 8.81 9 8.77
11 8.33 11 9.26
14 9.96 14 8.1
6 7.24 6 6.13
4 4.26 4 3.10
12 10.84 12 9.13
7 4.82 7 7.26
5 5.68 5 4.74
Average 9 7.501 9 7.501
Standard Deviation 3.317 2.032 3.317 2.032

2
Anscombe, F. J., “The Validity of Comparative Experiments,” Journal of the Royal Statistical Society, Vol. 11,
No. 3, 1948, pp. 181–211.
 1-2 Why Visualize Data? 7

FIGURE 1.2 Anscombe’s Data Displayed Graphically

Data Set 1
Y
12

10

4 y = 0.5x + 3.00
R² = 0.67
2

0
0 2 4 6 8 10 12 14 16
X
(a)
Anscombe
Data Set 2
Y
12

10

4 y = 0.5x + 3.00
R² = 0.67
2

0
0 2 4 6 8 10 12 14 16
X
(b)

Data Visualization for Explanation

Data visualization is also important for explaining relationships found in data and for
explaining the results of predictive and prescriptive models. More generally, data visual-
ization is helpful in communicating with your audience and ensuring that your audience
understands and focuses on your intended message.
Let us consider the article, “Check Out the Culture Before a New Job,” which appeared
in The Wall Street Journal.3 The article discusses the importance of finding a good cultural
fit when seeking a new job. Difficulty in understanding a corporate culture or misalignment
with that culture can lead to job dissatisfaction. Figure 1.3 is a re-creation of a bar chart
that appeared in this article. A bar chart shows a summary of categorical data using the
length of horizontal bars to display the magnitude of a quantitative variable.
The chart shown in Figure 1.3 shows the percentage of the 10,002 survey respon-
dents who listed a factor as the most important in seeking a job. Notice that our
attention is drawn to the dark blue bar, which is “Company culture” (the focus of the

3
Lublin, J. S. “Check Out the Culture Before a New Job,” The Wall Street Journal, January 16, 2020.
8 Chapter 1 Introduction

article). We immediately see that only “Salary and bonus” is more frequently cited
than “Company culture.” When you first glance at the chart, the message that is com-
The effective use of color is
municated is that corporate culture is the second most important factor cited by job
discussed in more detail in seekers. And as a reader, based on that message, you then decide whether the article is
Chapter 4. worth reading.

FIGURE 1.3 A Bar Chart of Survey Results of Job Seekers

What matters most to you when deciding which job to take next?

Salary and Bonus 24%

Company Culture 22%

Location 13%

Flexible Schedule 11%

Day-to-day Work 11%

Industry 8%

Job Title 6%

Health Care Benefits 5%

1-3 Types of Data

Different types of charts are more effective than others for certain types of data. For that
reason, let us discuss the different types of data you might encounter.
The Dow Jones Industrial Table 1.3 contains information on the 30 companies that make up the Dow Jones
Average is a stock market
Industrial Index (DJI). The table contains the company name, the stock symbol, the indus-
index. It was created in 1896
by Charles Dow. The 30
try type, the share price, and the volume (number of shares traded). We will use the data
companies that are included in contained in Table 1.3 to facilitate our discussion.
The Dow change periodically
to reflect changes in major
corporations in the United Quantitative and Categorical Data
States.
Quantitative data are data for which numerical values are used to indicate magnitude,
such as how many or how much. Arithmetic operations, such as addition, subtraction,
multiplication, and division, can be performed on quantitative data. For instance,
we can sum the values for Volume in Table 1.3 to calculate a total volume of all
shares traded by companies included in the Dow, because Volume is a quantitative
variable.
Categorical data are data for which categories of like items are identified by labels or
names. Arithmetic operations cannot be performed on categorical data. We can summarize
categorical data by counting the number of observations or computing the proportions of
observations in each category. For instance, the data in the Industry column in Table 1.3
are categorical. We can count the number of companies in the Dow that are, for example,
in the food industry. Table 1.3 shows two companies in the food industry: Coca-Cola and
McDonald’s. However, we cannot perform arithmetic operations directly on the data in the
Industry column.
1-3 Types of Data 9

TABLE 1.3  ata for the Dow Jones Industrial Index Companies
D
(April 3, 2020)
Company Symbol Industry Share Price ($) Volume
Apple Inc. AAPL Technology 241.41 32,470,017
American Express AXP Financial Services 73.6 9,902,194
Boeing BA Manufacturing 124.52 36,489,379
Caterpillar Inc. CAT Manufacturing 114.67 4,803,174
Cisco Systems CSCO Technology 39.06 21,235,157
Chevron CVX Petroleum 75.11 14,317,998
Disney DIS Entertainment 93.88 14,592,062
Goldman Sachs GS Financial Services 146.93 2,773,298
Home Depot, Inc. HD Retailing 178.7 6,762,357
IBM IBM Technology 106.34 3,909,196
Intel Corporation INTC Technology 54.13 23,906,062
Johnson & Johnson JNJ Pharmaceutical 134.17 9,409,033
JPMorgan Chase JPM Financial Services 84.05 20,363,095
Coca-Cola KO Food 43.83 13,294,556
McDonald’s MCD Food 160.33 4,361,094
3M Company MMM Conglomerate 133.79 3,461,642
Merck & Co. MRK Pharmaceutical 76.25 9,181,539
Microsoft MSFT Technology 153.83 41,243,284
Nike NKE Apparel 78.86 8,297,443
Pfizer PFE Pharmaceutical 33.64 30,306,371
Procter & Gamble PG Consumer Goods 115.08 7,520,086
Travelers TRV Financial Services 93.89 1,595,000
UnitedHealth Group UNH Healthcare 229.49 4,356,992
Raytheon UTX Conglomerate 86.01 13,203,254
Visa V Financial Services 151.85 11,649,519
Verizon VZ Telecommunication 54.7 16,304,703
Walgreens WBA Retailing 40.72 6,489,129
Walmart WMT Retailing 119.48 9,390,287
Exxon Mobil XOM Petroleum 39.21 48,094,821

Cross-Sectional and Time Series Data

We distinguish between cross-sectional data and times series data. Cross-sectional data
are collected from several entities at the same or approximately the same point in time. The
data in Table 1.3 are cross-sectional because they describe the 30 companies that comprise
the Dow at the same point in time (April 2020).
Time series data are data collected over several points in time (minutes, hours,
days, months, years, etc.). Graphs of time series data are frequently found in business,
economic, and science publications. Such graphs help analysts understand what hap-
pened in the past, identify trends over time, and project future levels for the time series.
10 Chapter 1 Introduction

For example, the graph of the time series in Figure 1.4 shows the DJI value from January
2010 to April 2020. The graph shows the upward trend of the DJI value from 2010
to 2020, when there was a steep decline in value due to the economic impact of the
COVID-19 pandemic.

Big Data
There is no universally accepted definition of big data. However, probably the most general
definition of big data is any set of data that is too large or too complex to be handled by
standard data-processing techniques using a typical desktop computer. People refer to the
four Vs of big data:
●● volume—the amount of data generated
●● velocity—the speed at which the data are generated
●● variety—the diversity in types and structures of data generated

●● veracity—the reliability of the data generated

Volume and velocity can pose a challenge for processing analytics, including data visual-
ization. Special data management software such as Hadoop and higher capacity hardware
(increased server or cloud computing) may be required. The variety of the data is handled
by converting video, voice, and text data to numerical data, to which we can then apply
standard data visualization techniques.
In summary, the type of data you have will influence the type of graph you should use to
convey your message. The zoo attendance data in Figure 1.1 are time series data. We used
a column chart in Figure 1.1 because the numbers are the total attendance for each month,
and we wanted to compare the attendance by month. The height of the columns allows us
to easily compare attendance by month. Contrast Figure 1.1 with Figure 1.4, which is also
time series data. Here we have the value of the Dow Jones Index. These data are a snapshot
of the current value of the DJI on the first trading day of each month. They provide what is

FIGURE 1.4 Dow Jones Index Values from January 2010 to April 2020

DJI Value
30,000

25,000

20,000

15,000
DJI
10,000

5,000

0
11

13
12
10

20
14

18

19
16
15

17
20

20
20
20

20
20

20

20
20
20

20
1/

1/
1/
1/

1/
1/

1/

1/
1/
1/

1/
1/

1/
1/
1/

1/
1/

1/

1/
1/
1/

1/
 1-4 Data Visualization in Practice 11

How to select an effective essentially a time path of the value, and so we use a line graph to emphasize the continuity
chart type is discussed in more
of time.
detail in Chapter 2.

1-4 Data Visualization in Practice

Data visualization is used to explore and explain data and to guide decision making in
all areas of business and science. Even the most analytically advanced companies such
as Google, Uber, and Amazon rely heavily on data visualization. Consumer goods giant
Procter & Gamble (P&G), the maker of household brands such as Tide, Pampers, Crest,
and Swiffer, has invested heavily in analytics, including data visualization. P&G has
built what it calls the Business Sphere™ in more than 50 of its sites around the world.
The Business Sphere is a conference room with technology for displaying data visual-
izations on its walls. The Business Sphere displays data and information P&G executives
and managers can use to make better-informed decisions. Let us briefly discuss some
ways in which the functional areas of business, engineering, science, and sports use data
visualization.

Accounting
Accounting is a data-driven profession. Accountants prepare financial statements and
examine financial statements for accuracy and conformance to legal regulations and best
practices, including reporting required for tax purposes. Data visualization is a part of
every accountant’s tool kit. Data visualization is used to detect outliers that could be an
indication of a data error or fraud. As an example of data visualization in accounting, let us
consider Benford’s Law.
Benfords Law, also known as the First-Digit Law, gives the expected probability that
the first digit of a reported number takes on the values one through nine, based on many
real-life numerical data sets such as company expense accounts. A column chart displaying
Benford’s Law is shown in Figure 1.5. We have rounded the probabilities to four digits. We
see, for example, that the probability of the first digit being a 1 is 0.3010. The probability
of the first digit being a 2 is 0.1761, and so forth.

FIGURE 1.5 A Column Chart Showing Benford’s Law

Benford’s Law: The Probability of the First Digit

0.3010

0.1761

0.1249
0.0969
0.0792
0.0669 0.0580 0.0512 0.0458

1 2 3 4 5 6 7 8 9
First Digit
12 Chapter 1 Introduction

Benford’s Law can be used to detect fraud. If the first digits of numbers in a data set
do not conform to Bedford’s Law, then further investigation of fraud may be warranted.
Consider the accounts payable (money owed the company) for Tucker Software. Figure 1.6
is a clustered column chart (also known as a side-by-side column chart). A clustered
column chart is a column chart that shows multiple variables of interest on the same
chart, with the different variables usually denoted by different colors or shades of a color.
In Figure 1.6, the two variables are Benford’s Law probability and the first digit data for a
random sample of 500 of Tucker’s accounts payable entries. The frequency of occurrence
in the data is used to estimate the probability of the first digit for all of Tucker’s accounts
payable entries. It appears that there are an inordinate number of first digits of 5 and 9 and
a lower than expected number of first digits of 1. These might warrant further investigation
by Tucker’s auditors.

FIGURE 1.6 A Clustered Column Chart Showing Benford’s Law versus

Tucker Software’s Accounts Payable Entries

Benford’s Law versus Tucker Software Accounts Payable

Probability
0.35

0.30 Benford Tucker

0.25

0.20

0.15

0.10

0.05

0.00
1 2 3 4 5 6 7 8 9
First Digit

Finance
Like accounting, the area of business known as finance is numerical and data-driven.
Finance is the area of business concerned with investing. Financial analysts, also known
as “quants,” use massive amounts of financial data to decide when to buy and sell certain
stocks, bonds, and other financial instruments. Data visualization is useful in finance for
recognizing trends, assessing risk, and tracking actual versus forecasted values of metrics
of concern.
Yahoo! Finance and other websites allow you to download daily stock price data. As an
example, the file Verizon has five days of stock prices for telecommunications company
We discuss High-Low-Close Verizon Wireless. Each of the five observations includes the date, the high share price for
Stock charts in more detail in
that date, the low share price for that day, and the closing share price for that day. Excel has
Chapter 2.
several charts designed for tracking stock performance with such data. Figure 1.7 displays
1-4 Data Visualization in Practice 13

these data in a high-low-close stock chart, a chart that shows the high value, low value,
and closing value of the price of a share of stock over time. For each date shown, the bar
indicates the range of the stock price per share on that day, and the labelled point on the
bar indicates closing price per share for that day. The chart shows how the closing price is
changing over time and the volatility of the price on each day.

FIGURE 1.7 A High-Low-Close Stock Chart for Verizon Wireless

Verizon Wireless Stock Price per Share Performance

Price per Share ($) Close
59.50

59.00

58.50
58.13
58.00 57.99 57.93

57.50 57.59

57.00
56.82
56.50

56.00

55.50
20-Apr 21-Apr 22-Apr 23-Apr 24-Apr

Human Resource Management

Human resource management (HRM) is the part of an organization that focuses on an orga-
nization’s recruitment, training, and retention of employees. With the increased use of ana-
lytics in business, HRM has become much more data-driven. Indeed, HRM is sometimes
now referred to as “people analytics.” HRM professionals use data and analytical models to
form high-performing teams, monitor productivity and employee performance, and ensure
diversity of the workforce. Data visualization is an important component of HRM, as HRM
professionals use data dashboards to monitor relevant data supporting their goal of having
a high-performing workforce.
A key interest of HRM professionals is employee churn, or turnover in an organiza-
tion’s workforce. When employees leave and others are hired, there is often a loss of pro-
ductivity as positions go unfilled. Also, new employees typically have a training period
and then must gain experience, which means employees will not be fully productive at
the beginning of their tenure with the company. Figure 1.8, a stacked column chart, is an
example of a visual display of employee turnover. It shows gains and losses of employees
by month. A stacked column chart is a column chart that shows part-to-whole compari-
sons, either over time or across categories. Different colors or shades of color are used to
denote the different parts of the whole within a column. In Figure 1.8, gains in employees
(new hires) are represented by positive numbers in darker blue and losses (people leaving
the company) are presented as negative numbers and lighter blue bars. We see that January
and July–October are the months during which the greatest numbers of employees left the
company, and the months with the highest numbers of new hires are April through June.
14 Chapter 1 Introduction

Visualizations like Figure 1.8 can be helpful in better understanding and managing work-
force fluctuations.

FIGURE 1.8 A Stacked Column Chart of Employee Turnover by Month

Number of Employees
60

50 Gains Losses

40

30

20

10

–10

–20

–30
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month

Marketing
Marketing is one of the most popular application areas of analytics. Analytics \is used
for optimal pricing, markdown pricing for seasonal goods, and optimal allocation of
marketing budget. Sentiment analysis using text data such as tweets, social networks to
determine influence, and website analytics for understanding website traffic and sales,
are just a few examples of how data visualization can be used to support more effective
marketing.
Let us consider a software company’s website effectiveness. Figure 1.9 shows a funnel
chart of the conversion of website visitors to subscribers and then to renewal customers.
Funnel charts are discussed in A funnel chart is a chart that shows the progression of a numerical variable for various
more detail in Chapter 2.
categories from larger to smaller values. In Figure 1.9, at the top of the funnel, we track
100% of the first-time visitors to the website over some period of time, for example, a
six-month period. The funnel chart shows that of those original visitors, 74% return to
the website one or more times after their initial visit. Sixty-one percent of the first-time
visitors downloaded a 30-day trial version of the software, 47% eventually contacted
support services, 28% purchased a one-year subscription to the software, and 17% even-
tually renewed their subscription. This type of funnel chart can be used to compare the
conversion effectiveness of different website configurations, the use of bots, or changes in
support services.

Operations
Like marketing, analytics is used heavily in managing the operations function of busi-
ness. Operations management is concerned with the management of the production and
1-4 Data Visualization in Practice 15

FIGURE 1.9 A Funnel Chart of Website Conversions for a Software Company

Visited the Website 100%

Returned to the Website 74%

Downloaded a Trial Version 61%

Contacted Support 47%

Subscribed 28%

Renewed 17%

distribution of goods and services. It includes responsibility for planning and scheduling,
inventory planning, demand forecasting, and supply chain optimization. Figure 1.10
shows time series data for monthly unit sales for a product (measured in thousands of
units sold). Each period corresponds to one month. So that a cost-effective produc-
tion schedule can be developed, an operations manager might have responsibility for

FIGURE 1.10 Time Series Data for Units Sales of a Product

Sales (1000s units)

3000

2500

2000

1500

1000

500

0
0 5 10 15 20 25 30 35 40
Month
16 Chapter 1 Introduction

forecasting the monthly unit sales for next twelve months (periods 37–48). In looking at
the time series data in Figure 1.10, it appears that there is a repeating pattern and units
sold might also be increasing slightly over time. The operations manager can use these
observations to help guide the forecasting techniques to test to arrive at reasonable fore-
casts for periods 37–48.

Engineering
Engineering relies heavily on mathematics and data. Hence, data visualization is an impor-
tant technique in every engineer’s toolkit. For example, industrial engineers monitor the
production process to ensure that it is “in control” or operating as expected. A control
chart is a graphical display that is used to help determine if a production process is in
control or out of control. A variable of interest is plotted over time relative to lower and
upper control limits. Consider the control chart for the production of 10-pound bags of dog
food shown in Figure 1.11. Every minute, a bag is diverted from the line and automatically
weighed. The result is plotted along with lower and upper control limits obtained statisti-
cally from historical data. When the points are between the lower and upper control limits,
the process is considered to be in control. When points begin to appear outside the control
limits with some regularity and/or when large swings start to appear as in Figure 1.11, this
is a signal to inspect the process and make any necessary corrections.

FIGURE 1.11 A Quality Control Chart for Dog Food Production

Weight (pounds)
10.10
10.08
10.06 Upper Control Limit
10.04
10.02
10.00
9.98
9.96 Lower Control Limit
9.94
9.92
9.90
1 3 5 7 9 11 13 15
Minute

Sciences
The natural and social sciences rely heavily on the analysis of data and data visualization
for exploring data and explaining the results of analysis. In the natural sciences, data are
often geographic, so maps are used frequently. For example, the weather, pandemic hot
spots, and species distributions can be represented on a geographic map. Geographic maps
are not only used to display data, but also to display the results of predictive models. An
example of this is shown in Figure 1.12. Predicting the path a hurricane will follow is a
1-4 Data Visualization in Practice 17

FIGURE 1.12 A Spaghetti Chart of Hurricane Paths from Multiple Predictive

Models

complicated problem. Numerous models, each with its own set of influencing variables
(also known as model features), yield different predictions. Displaying the results of each
model on a map gives a sense of the uncertainty in predicted paths across all models and
expands the alert to a broader range of the population than relying on a single model.
Because the multiple paths resemble pieces of spaghetti, this type of map is sometimes
referred to as a “spaghetti chart.” More generally, a spaghetti chart is a chart depicting
possible flows through a system using a line for each possible path.

Sports
The use of analytics in sports has gained considerable notoriety since 2003, when
renowned author Michael Lewis published his book Moneyball. Lewis’s book tells how
the Oakland Athletics used an analytical approach for player evaluation to assemble a
competitive team using a limited budget. The use of analytics for player evaluation and on-
field strategy is now common throughout professional sports. Data visualization is a key
component of how analytics is applied in sports. It is common for coaches to have tablet
computers on the sideline that they use to make real-time decisions such as calling plays
and making player substitutions.
Figure 1.13 shows an example of how data visualization is used in basketball. A shot
chart is a chart that displays the location of the shots attempted by a player during a
basketball game with different symbols or colors indicating successful and unsuccess-
ful shots. Figure 1.12 shows shot attempts by NBA player Chris Paul, with a blue dot
indicating a successful shot and a orange x indicating a missed shot (source: Basketball-
Reference.com). Other NBA teams can utilize this chart to help devise strategies for
defending Chris Paul.
18 Chapter 1 Introduction

FIGURE 1.13 A Shot Chart for NBA Player Chris Paul

No t e s 1 C o mm e n t s

Chart is considered a more general term than graph. For (a line chart). In this text, we use the terms chart and graph
example, charts encompass maps, bar charts, etc., but graphs interchangeably.
generally refer to a chart of the type shown in Figure 1.4

S U M M A RY

This introductory chapter began with a discussion of analytics, the scientific process of
transforming data into insights for making better decisions. We discussed the three types of
analytics: descriptive, predictive, and prescriptive. Descriptive analytics describes what has
happened and includes tools such as reports, data visualization, data dashboards, descrip-
tive statistics, and some data-mining techniques. Predictive analytics consists of techniques
that use past data to predict future events or understand the relationships between variables.
These techniques include regression, data mining, forecasting, and simulation. Prescriptive
analytics uses input data to suggest a decision or course of action. This class of analytical
techniques includes rule-based models, simulation, decision analysis, and optimization.
Descriptive and predictive analytics can help us better understand the uncertainty and risk
associated with our decision alternatives.
This text focuses on descriptive analytics, and in particular on data visualization. Data
visualization can be used for exploring data and for explaining data and the output of anal-
yses. We explore data to more easily identify patterns, recognize anomalies or irregularities
in the data, and better understand relationships between variables. Visually displaying data
enhances our ability to identify these characteristics of data. Often we put various charts
and tables of several related variables into a single display called a data dashboard. Data
dashboards are collections of tables, charts, maps, and summary statistics that are updated
Glossary 19

as new data become available. Many organizations and businesses use data dashboards to
explore and monitor performance data such as inventory levels, sales, and the quality of
production.
We also use data visualization for explaining data and the results of data analyses. As
business becomes more data-driven, it is increasingly important to be able to influence
decision making by telling a compelling data-driven story with data visualization. Much
of the rest of this text is devoted to how to visualize data to clearly convey a compelling
message.
The type of chart, graph, or table to use depends on the type of data you have and
your intended message. Therefore, we discussed the different types of data. Quantitative
data are numerical values used to indicate magnitude, such as how many or how much.
Arithmetic operations, such as addition and subtraction, can be performed on quantitative
data. Categorical data are data for which categories of like items are identified by labels
or names. Arithmetic operations cannot be performed on categorical data. Cross-sectional
data are collected from several entities at the same or approximately the same point in
time, whereas time series data are collected on a single variable at several points in time.
Big data is any set of data that is too large or complex to be handled by typical data-pro-
cessing techniques using a typical desktop computer. Big data includes text, audio, and
video data.
We concluded the chapter with a discussion of applications of data visualization in
accounting, finance, human resource management, marketing, operations, engineering,
science, and sports, and we provided an example for each area. Each of the remaining
chapters of this text will begin with a real-world application of a data visualization. Each
Data Visualization Makeover is a real visualization we discuss and then improve by apply-
ing the principles of the chapter.

G L O S S A RY

Analytics The scientific process of transforming data into insights for making better
decisions.
Bar chart A chart that shows a summary of categorical data using the length of horizontal
bars to display the magnitude of a quantitative variable.
Big data Any set of data that is too large or complex to be handled by standard data-
processing techniques using a typical desktop computer. Big data includes text, audio, and
video data.
Categorical data Data for which categories of like items are identified by labels or names.
Arithmetic operations cannot be performed on categorical data.
Clustered column chart A column chart showing multiple variables of interest on the
same chart, the different variables usually denoted by different colors or shades of a color
with the columns side by side.
Column chart A chart that shows numerical data by the height of a column for a variety of
categories or time periods.
Control chart A graphical display in which a variable of interest is plotted over time
relative to lower and upper control limits.
Cross-sectional data Data collected from several entities at the same or approximately the
same point in time.
Data dashboard A data visualization tool that gives multiple outputs and may update in
real time.
Data visualization The graphical representation of data and information using displays
such as charts, graphs, and maps.
Descriptive analytics The set of analytical tools that describe what has happened.
Funnel chart A chart that shows the progression of a numerical variable to typically
smaller values through a process, for example, the percentage of website visitors who
ultimately result in a sale.
20 Chapter 1 Introduction

High-low-close stock chart A chart that shows three numerical values: high value, low
value, and closing value for the price of a share of stock over time.
Predictive analytics Techniques that use models constructed from past data to predict
future events or better understand the relationships between variables.
Prescriptive analytics Mathematical or logical models that suggest a decision or course of
action.
Quantitative data Data for which numerical values are used to indicate magnitude,
such as how many or how much. Arithmetic operations, such as addition, subtraction,
multiplication, and division, can be performed on quantitative data.
Scatter chart A graphical presentation of the relationship between two quantitative
variables. One variable is shown on the horizontal axis and the other is shown on the
vertical axis.
Shot chart A chart that displays the location of shots attempted by a basketball player
during a basketball game with different symbols or colors indicating successful and
unsuccessful shots.
Spaghetti chart A chart depicting possible flows through a system using a line for each
possible path.
Time series data Data collected over several points in time (minutes, hours, days, months,
years, etc.).

P R O B L E M S

1. Types of Analytics. Indicate which type of analytics (descriptive, predictive, or pre-

scriptive analytics) each of the following represents. LO 1
a. a data dashboard
b. a model that finds the production schedule that minimizes overtime
c. a model that forecasts sales for the next quarter
d. a bar chart
e. a model that allocates your financial investments to achieve your financial goal
2. Transportation Planning. An analytics professional is asked to plan the shipment of a
product for the next quarter. She employs the following process:
Step 1. For each of the 12 distribution centers, she plots the quarterly demand for the
product over the last three years.
Step 2. Based on the plot for each distribution center, she develops a forecasting
model to forecast demand for next quarter for each distribution center.
Step 3. She takes the forecast for next quarter for each distribution center and inputs
those forecasts, along with the capacities of the company’s four factories and
transportation rates from each factory to each distribution center, into an opti-
mization model. The optimization model suggests a shipping plan that min-
imizes the cost of how to satisfy the forecasted demand from the company’s
four different factories to the distribution centers.
Describe the type of analytics being utilized in each of the three steps outlined above.
LO 1
3. Wall Street Journal Subscriber Characteristics. A Wall Street Journal subscriber
survey asked a series of questions about subscriber characteristics and interests. State
whether each of the following questions provides categorical or quantitative data. LO 2
a. What is your age?
b. Are you male or female?
c. When did you first start reading the WSJ? High school, college, early career, midca-
reer, late career, or retirement?
d. How long have you been in your present job or position?
e. What type of vehicle are you considering for your next purchase? Nine response
categories for this question include sedan, sports car, SUV, minivan, and so on.
Problems 21

4. Comparing Smartwatches. Consumer Reports provides product evaluations for its

subscribers. The following table shows data from Consumer Reports for five smart-
watches on the following characteristics:
Overall Score—a score awarded for a variety of performance factors
Price—the retail price
Recommended—does Consumer Reports recommend purchasing the smartwatch based
on performance and strengths?
Best Buy—if Consumer Reports recommends purchasing the smartwatch, does it also
consider it a “best buy” based on a blend of performance and value?

Make Overall Score Recommended Best Buy Price

Apple Watch Series 5 84 Yes No $395
Fitbit Versa 2 78 Yes Yes $200
Garmin Venu 77 Yes No $350
Fitbit Versa Lite 65 No No $100

For each of the four pieces of data, indicate whether the data are quantitative or cate-
gorical and whether the data are cross-sectional or time series. LO 2
5. House Price and Square Footage. Suppose we want to better understand the relation-
ship between house price and square footage of the house, and we have collected house
price and square footage for 75 houses in a particular neighborhood of Cincinnati,
Ohio, from the Zillow website on January 3, 2021. LO 2, 3
a. Are these data quantitative or categorical?
b. Are these data cross-sectional or times series?
c. Which of the following type of chart would provide the best display of these data?
Explain your answer.
i. Bar chart
ii. Column chart
iii. Scatter chart
6. Netflix Subscribers. The following chart displays the total number of Netflix sub-
scribers from 2010 to 2019. LO 1, 2, 3
a. Are these data quantitative or categorical?
b. Are these data cross-sectional or time series?
c. What type of chart is this?

Netflix Subscribers (millions)

167.1

139.3

110.6
89.1
74.8
57.4
44.4
33.3
26.3
20.0

2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
Year

7. U.S. Netflix Subscribers. Refer to the previous problem. Suppose that in addition
to the total number of Netflix subscribers, we have the number of those subscribers
by year for the years 2010–2019 who live in the United States. Our message is to
22 Chapter 1 Introduction

emphasize how much of the growth is coming from the United States. Which of the
following types of charts would best display the data? Explain your answer. LO 2, 3
i. Bar chart
ii. Clustered column chart
iii. Stacked column chart
iv. Stock chart
8. How Data Scientists Spend Their Day. The Wall Street Journal reported the results
of a survey of data scientists. The survey asked the data scientists how they spend their
time. The following chart shows the percentage of respondents who answered less than
five hours per week or at least five hours per week for the amount of time they spend
on exploring data and on presenting analyses. LO 2, 3, 4

What Data Scientists Do: Exploring versus Presenting

74%
Presenting Analysis
26%

Less than five hours per week At least five hours per week

42%
Exploring Data
58%

a. Are these data quantitative or categorical?

b. Are these data cross-sectional or time series?
c. What type of chart is this?
d. What conclusions can you make based on this chart?
9. Industries in the Dow Jones Industrial Index. Refer to the data on the Dow Jones
Industrial Index given in Table 1.3. The following chart displays the number of compa-
nies in each industry that make up this index. LO 3
a. What type of chart is this?
b. Which industry has the highest number of companies in the Dow Jones Industrial
Index?

Number of Companies by Industry

Technology 5
Financial Services 5
Retailing 3
Pharmaceutical 3
Petroleum 2
Manufacturing 2
Food 2
Conglomerate 2
Telecommunication 1
Healthcare 1
Entertainment 1
Consumer Goods 1
Apparel 1
Problems 23

10. Job Factors. The following chart is based on the same data used to construct
Figure 1.3. The data are percentages of respondents to a survey who listed various
factors as most important when making a job decision. LO 3, 4
a. What type of chart is this?
b. What is the fifth most-cited factor?

What matters most to you when deciding which job to take next?
24%
22%

13%
11% 11%

8%
6%
5%

Salary and Company Location Day-to-day Flexible Industry Job Title Health Care
Bonus Culture Work Schedule Benefits

11. Retirement Financial Concerns. The results of the American Institute of Certified
Public Accountants’ Personal Financial Planning Trends Survey indicated 48% of
clients had concerns about outliving their money. The top reasons for these concerns
and the percentage of respondents who cited the reason were as follows. LO 3, 4
Concerns for Retirement

Health-care Costs 77%

Stock Market Fluctuations 53%

Unexpected Costs 50%

Lifestyle Changes 42%

Possibility of Being a Financial Burden 22%

Desire to Leave an Inheritance 21%

a. What type of chart is this?

b. Only 48% of the survey respondents had financial concerns about retirement
(outliving their money). What percentage of the total people surveyed had retire-
ment health-care cost concerns?
12. Master’s Degree Program Recruiting. The recruiting process for a full-time master’s
program in data science consists of the following steps. The program director obtains
email addresses of undergraduate seniors who have taken the Graduate Record Exam
(GRE) and expressed an interest in data science. An email inviting the students to an
24 Chapter 1 Introduction

online information session is sent. At the information session, faculty discuss the pro-
gram and answer questions. Students apply through a web portal. An admissions com-
mittee makes an offer of admission (or not) along with any financial aid. If the person
is admitted, the person either accepts or rejects the offer. Consider the following chart.
LO 3, 4

Master’s Degree in Data Science Recruiting

Email 100%

Information Session 64%

Applied for Admission 47%

Admitted 25%

Enrolled 21%

a. What type of chart is this?

b. Which of the following is the correct interpretation of the 21% for Enrolled?
i. Of those who were sent an email, 21% enrolled.
ii. Of those who were admitted, 21% enrolled.
iii. Of those who applied for admission, 21% enrolled.
iv. None of the above
13. Chemical Process Control. The following chart is a quality control chart of the tem-
perature of a chemical manufacturing process. What observations can you make about
the process? LO 3

Temperature (degrees Fahrenheit)

97.00
Upper Control Limit
96.80

96.60

96.40

96.20

96.00 Lower Control Limit

95.80

95.60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Hour
Problems 25

14. Buying a Used Car. The following chart shows data for a sample of 18 used cars of
the same brand, model, and year. LO 2, 3, 4
a. Are these data quantitative or categorical?
b. What type of chart is this?
c. How might you use this chart to find a used car to purchase?

Price ($1000s)
16.0

14.0

12.0

10.0

8.0

6.0

4.0

2.0

0.0
0 20 40 60 80 100 120
Mileage (1000s)

15. Tracking Stock Prices. The following high-low-close stock chart gives the stock price
for Exxon Mobile Corporation over a 12-month period. The data are the low, high, and
closing price per share on the first trading day of the month. What can you say about
the stock price and volatility of the stock price over this 12-month period? LO 3

Price ($ per share)

Close
90.00

80.00
76.63
74.36
70.00 70.61
68.48 67.57 68.13 69.78
62.12
60.00

50.00 51.44
46.47 46.18
40.00
37.97

30.00

20.00

10.00

0.00
1 2 3 4 5 6 7 8 9 10 11 12
Month
Chapter 2
Selecting a Chart Type
CONTENTS

DATA VISUALIZATION MAKEOVER 2-5 MAPS

The New York City Comptroller Geographic Maps
Heat Maps
2-1 DEFINING THE GOAL OF YOUR DATA Treemaps
VISUALIZATION 2-6 WHEN TO USE TABLES
Selecting an Appropriate Chart Tables versus Charts
2-2 CREATING AND EDITING CHARTS IN EXCEL 2-7 OTHER SPECIALIZED CHARTS
Creating a Chart in Excel Waterfall Charts
Editing a Chart in Excel Stock Charts
Funnel Charts
2-3 SCATTER CHARTS AND BUBBLE CHARTS
Scatter Charts 2-8 A SUMMARY GUIDE TO CHART SELECTION
Bubble Charts Guidelines for Selecting a Chart
Some Charts to Avoid
2-4 LINE CHARTS, COLUMN CHARTS, AND BAR
Excel’s Recommended Charts Tool
CHARTS
Line Charts SUMMARY
Column Charts
GLOSSARY
Bar Charts
PROBLEMS

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Create charts and graphs using Excel LO 4 Interpret insights from charts and graphs

LO 2 Modify charts and graphs using Excel LO 5 Recognize which chart types should be avoided
and explain why
LO 3 Identify an appropriate chart type for a given goal
and data type
 Data Visualization Makeover 27

D ata V is u ali z ation M a k eover

The New York City Comptroller

The New York City (NYC) comptroller’s office has to each of ten spending categories. For each of the ten
roughly 800 employees. Accountants, economists, categories, it also expresses the respective spending
engineers, investment analysts, information technol- allocation as a percentage of the total $70.24 billion
ogy support, and administrative support all support dollar budget for fiscal years 2020–2023.1
the mission of the NYC Comptroller, namely, to ensure The audience for the report is the public and most
the fiscal health of New York City. The Comptroller’s likely New York City residents who pay taxes and are
office is responsible for: auditing performance and interested in how the city allocates its budget. people
efficiency; ensuring integrity in city contracting; with a passion for a particular cause might also be
managing assets to protect pensions; resolving claims interested in this chart. For example, an advocate for
against the city and risk management; managing city parks and recreation might want to know how much
bonds; enforcing labor rights; and promoting fiscal money has been allocated to that cause and how
health and a sound budget for New York City. much it is relative to other spending categories.
In its work, the comptroller’s office generates a vari- Figure 2.2 is a horizontal bar chart that displays the
ety of annual reports including the Annual Audit Report, budget allocation amounts. Most data visualization
Annual Analysis of NYC Agency Contracts, Annual experts suggest that pie charts should be avoided in
Claims Report, and the Annual Report on Capital Debt favor of bar charts. There are several reasons for this.
and Obligations. The pie chart in Figure 2.1 is from First, science has shown that we are better at assessing
the Annual Report on Capital Debt and Obligations. differences in length than angle and area. Glancing at
It shows the amounts (in millions of dollars) allocated the pie chart in Figure 2.1 and comparing the Other

FIGURE 2.1 A Pie Chart Showing the Allocation of New York City Funds

Parks
$3,876 Other City
5% Operations
$7,655
Hospitals 11% Education/CUNY
$1,257 $14,645
2% 21%

Admin. of Justice
$7,393
11% Environmental
Protection
$10,667
15%

Citywide
Equipment DOT & Mass
Housing & Transit
$4,019 Economic
6% $9,486
Development 14%
Computer $10,771
Equipment 15%
$473
1%

Source: NYC office of Management and Budget, FY 2020 Adopted Capital Commitment Plan, October 2019.

1
Note that because of rounding, percentages do not add up to 100%.
(Continued)
28 Chapter 2 Selecting a Chart Type

Figure 2.2 A Bar Chart Showing the NYC Budget Allocation

Amount ($ millions) Allocated for New York City 2020–2023

Education/CUNY 14,645

Housing & Economic Development 10,771

Environmental Protection 10,667

DOT & Mass Transit 9,486

Other City Operations 7,655

Administration of Justice 7,393

Citywide Equipment 4,019

Parks 3,876

Hospitals 1,257

Computer Equipment 473

Total Budget ($ millions) 70,242

City Operations category with Administration of Justice distinguish the allocations by category. Third, we use a
category, it is difficult to tell which has a larger allocation. horizontal bar chart rather than a vertical bar chart so that
Indeed, the allocations for those two categories are very the category labels, a few of which are rather lengthy,
close. However, in Figure 2.2, because we have sorted are easier to read. Finally, we use the actual allocations
by amount allocated, we can see that the bar for Other and drop the percentages that appear in Figure 2.1. We
City Operations is longer (also, it appears higher up in could add the percentages, but it would make the
the list). Second, notice that in Figure 2.2, we no longer bar chart more crowded. Instead, we opt to use only
need to use different colors to distinguish categories. amounts because the bar lengths indicate the relative
color is not necessary to distinguish each category in allocations, and a reader interested in the percentages
this bar chart because the length of the bars is used to can calculate them from the given allocation amounts.

In this chapter, we discuss in more detail how to select the right chart type to most effec-
tively convey a message to your audience. In the case of the NYC Comptroller, we are
comparing the amounts allocated by category, so that the constituents of New York City
can compare the spending categories and assess for themselves the budget allocation. A bar
chart is appropriate for comparison.
There are numerous types of charts available, each designed for a purpose. Understanding
the different types of charts available and why some charts are more appropriate for a certain
purpose will make you a better data analyst and a better communicator with data. In this
chapter, we describe some of the most commonly used types of charts and when they should
be used. We also discuss some more advanced charts, as well as charts to be avoided.

2-1 Defining the Goal of Your Data Visualization

Selecting an Appropriate Chart
How do you choose an appropriate chart? If the goal of your chart is to explain, then the
answer to this question depends on the message you wish to convey to your audience. If
you are exploring data, the best chart type depends on the question you are asking and hope
 2-2 Creating and Editing Charts in Excel 29

to answer from the data. Also, the type of data you have may influence your chart selection.
A few of the more common goals for charts are to show the following:
●● Composition—Composition is what makes up the whole of an entity under consid-
eration. An example is the bar chart in Figure 2.2.
●● Ranking—Ranking is the relative order of items. Figure 2.2 is also an example of

ranking, because we have sorted the categories by bar length, which is proportional to
the amounts allocated.
●● Correlation/Relationship—Correlation is how two variables are related to one

another. An example of this is the relationship between average low temperature and
average annual snowfall for various cities in the United States.
●● Distribution—Distribution is how items are dispersed. An example of this is the

number of calls received by a call center in a day, measured on an hourly basis.

The type of data you have should also influence your chart selection. For example, a bar or
column chart is often an appropriate chart when we are summarizing data about categories.
Students’ letter grades in a college course are categories. For summarizing the number of
students earning each letter grade, a bar or column chart would be appropriate.
The different types of data are The relationship between two quantitative variables often makes a scatter chart an
discussed in Chapter 1.
appropriate choice. Bar charts, scatter charts, and line charts with the horizontal axis being
time, are often the best choice for time series data. If your data have a spatial component, a
geographic map might be a good choice.
Creating great data visualizations is a skill that is best learned by doing. Therefore, before
getting into more detail on the various types of charts and in what circumstances they are most
appropriate, we provide detailed instructions on how to create and edit charts in Excel.

2-2 Creating and Editing Charts in Excel

In this section, we discuss how to create and edit a chart in Excel. Let us begin with how
to create the chart of the zoo attendance data discussed in chapter 1. The data, zoo
attendance by month, are shown in Figure 2.3. We create a column chart using these
data in the following steps.

FIGURE 2.3 Data in File Zoo

Zoo
30 Chapter 2 Selecting a Chart Type

FIGURE 2.4 An Unedited Column Chart for the Zoo Data

Attendance
25000

20000

15000

10000

5000

0
Jan Feb Mar Apr May Jun July Aug Sept Oct Nov Dec

Creating a Chart in Excel

The following steps show how to create a column chart in Excel using the data in the file Zoo.
Step 1. Select cells A1:B13
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Column or Bar Chart button in the Charts group
When the list of column and bar charts subtypes appears, click the
Clustered Column button
The chart created in the preceding steps appears in Figure 2.4. We can improve the appear-
ance of the column chart in Figure 2.4 by following the steps below to delete the horizontal
grid lines, make the axes better defined add axis labels, and remove the border of the chart.
This will improve the chart by making it simpler and better-defined.

Editing a Chart in Excel

Here we give the step-by-step instructions on how to edit the chart shown in Figure 2.4,
and included in the file ZooChart, so that it appears as shown in Figure 2.5. These steps for
ZooChart
editing will be used to throughout the following chapters in this book to improve the for-
matting of most charts created in Excel.
Step 1 removes the horizontal grid lines and Step 2 edits the title.
The Chart Elements button is Step 1. Click anywhere on the chart then click on the border of the chart title and
not available in Mac versions press the Delete key
of Excel. See the NOTES +
Click the Chart Elements button
COMMENTS at the end of this
section for a description of
Deselect the check box for Gridlines and select the check box for Axis Titles
how to access these features Step 2. Click the Chart Title text box above the chart and replace “Chart Title” with
on a Mac. Zoo Attendance by Month
Inside the text box, highlight “Zoo Attendance by Month”
Click the Home tab on the Ribbon and in the Font group select
Calibri 16 pt Bold
 2-2 Creating and Editing Charts in Excel 31

FIGURE 2.5 A Column Chart of Zoo Attendance by Month

Zoo Attendance by Month

Attendance
25000

20000

15000

10000

5000

0
Jan Feb Mar Apr May Jun July Aug Sept Oct Nov Dec
Month

To change the position of the title, click on the border of the title box and
slide the text box to the left so that the title is above the vertical axis.
Steps 3–5 format the horizontal axis and axis labels.
Step 3. Double click any label of the horizontal axis
Step 4. When the Format Axis task pane appears, click the Fill & Line button
Click Line
Select Solid line
In the drop down to the right of Color, under Theme Colors, select Black
Step 5. Click the Home tab on the Ribbon and in the Font group select Calibri 10.5
Steps 6–8 format the vertical axis and axis labels.
Step 6. Double click any label of the vertical axis
Step 7. When the Format Axis task pane appears, click the Fill & Line button
Click Line
Select Solid line
In the drop down to the right of Color, under Theme Colors, select Black
Step 8. In the Format Axis task pane, click the Axis Options button
Click Tick Marks
Next to Major type, select Inside
Step 9. Click the Home tab on the Ribbon and in the Font group select Calibri 10.5
Steps 10–11 add and format axis titles.
Step 10. Select the horizontal axis title, place the cursor over the border of the text box
and drag it to the right to the end of the axis.
In the Font group, select Calibri 10.5 and click the Bold B button.
Type Month
Step 11. Select the vertical axis title, right click and select Format Axis Title and click
the Size & Properties button. Click Alignment and next to Text direction,
32 Chapter 2 Selecting a Chart Type

from the drop-down menu, select Horizontal. Place the cursor over the border
and drag it to the top of the vertical axis aligned above the axis labels.
In the Font group, select Calibri 10.5 and click the Bold B button.
Type Attendance
Steps 12–13 eliminate the border of the chart.
In Step 12, if you click inside Step 12. Click the Chart Area of the chart (anywhere outside of the rectangular area
the rectangular area delimited delimited by the horizontal and vertical axes of the chart)
by the horizontal and vertical
Step 13. In the Format Chart Area task pane, click Chart Options
axes, the Format Plot Area
task pane will be activated
Click the Fill & Line button
instead of the Format Chart Click Border
Area task pane. Click No line
These steps produce the chart shown in Figure 2.5. In later chapters, we will introduce
additional design elements that can be used to further improve charts.

N otes 1 C omments

4. The Chart Elements button offers a variety of features,

1. The Chart Elements button is not available in Excel for
including the ability to add or delete axes and axis titles, a
Mac. To access the features of Chart Elements in Excel for
chart title, data labels, gridlines, a legend, and a trendline.
Mac, double click the chart, and click the Chart Design tab,
5. An axis title can also be created using a text box by click-
and click Add Chart Element from the Chart Layouts group.
ing Insert on the Ribbon, clicking Text, and then selecting
2. When selecting data in Excel to populate a chart, the left-
where you want the axis title to appear.
most column of data is generally represented by the horizon-
6. The Chart Styles button allows you to change the
tal axis on the chart. The assignment of data to the horizontal
style of the chart type you have selected, as well the color
and vertical axes can be switched by right-clicking on the
scheme of the chart. The Chart Styles button is not avail-
chart, choosing Select Data Source, and on the Select Data
able in Excel for Mac. To access the features of Chart Styles
Source dialog box, click Switch Row/Column.
in Excel for Mac, double click the chart, click the Chart
3. To create a chart using nonadjacent columns of data in
Design tab, and click a style from the Chart Styles group.
Excel, select the leftmost column you wish to include in
To change the color scheme, click the Change Colors
the chart, press and hold down the control key (Ctrl), and
button in the Chart Styles group.
select the other columns of data you wish to include.

2-3 Scatter Charts and Bubble Charts

When exploring data, we are often interested in the relationship between two quantitative
variables. For example, we might be interested in the square footage of a house and the
cost of the house, or the age of a car and its annual maintenance cost. A scatter chart is a
graphical presentation of the relationship between two quantitative variables. One variable
is shown on the horizontal axis and the other is shown on the vertical axis, and a symbol is
used to plot ordered pairs of the quantitative variable values. A scatter chart is appropriate
for better understanding the relationship between two quantitative variables. As we shall
also see, a bubble chart is an appropriate chart when trying to show relationships with more
than two quantitative variables.

Scatter Charts
The file Snow contains the average low temperature in degrees Fahrenheit and the average
annual snowfall in inches for 51 major cities in the United States. A portion of the data
are shown in Figure 2.6. These averages are based on thirty years of data. Suppose we are
interested in the relationship between these two variables. Intuition tells us that the higher
the average low temperature the lower the average snowfall, but what is the nature of this
relationship?
 2-3 Scatter Charts and Bubble Charts 33

FIGURE 2.6 A Portion of the Data in File Snow

Snow

The data are plotted in Figure 2.7. This scatter chart is created using the following steps.
Step 1. Select cells C1:D52
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Scatter (X,Y) or Bubble Chart button in the Charts
group
When the list chart subtypes appears, click the Scatter button
Then edit the chart as outlined in Section 2-2.
Each point on the chart in Figure 2.7 represents a pair of numbers. In this case, we have a
pair of measurements for each of 51 cities. The measurements are average low temperature
in degrees Fahrenheit and average annual amount of snowfall in inches. We can see from
the chart that average annual amount of snowfall intuitively levels off at zero for warm-
weather cities.
Scatter charts are among the most useful charts for exploring pairs of quantitative data.
But, what if you wish to explore the relationships between more than two quantitative
variables? When exploring the relationships between three quantitative variables, a bubble
chart may be useful.

Bubble Charts
A bubble chart is a scatter chart that displays a third quantitative variable using different
sized dots, which we refer to as bubbles.
The file AirportData contains data on a sample of 15 airports. These data are shown
in Figure 2.8. For each airport, we have the following quantitative variables: average wait
time in the non-priority Transportation Security Authority (TSA) queue measured in min-
utes, the cheapest on-site daily rate for parking at the airport measured in dollars, and the
number of enplanements in a year (the number of passengers who board including trans-
AirportData fers) measured in millions of passengers.
34 Chapter 2 Selecting a Chart Type

FIGURE 2.7 A Scatter Chart for the Data in File Snow

Average Snowfall versus Average Low Temperature for 51 U.S. Cities

Average Annual Snowfall (inches)
120

100

80

60

40

20

0
0 10 20 30 40 50 60 70 80
Average Low Temperature (degrees Farenheit)

FIGURE 2.8 Data in file AirportData

The data are plotted as a bubble chart in Figure 2.9. This chart was created using the
following steps:
Step 1. Select cells B1:D16
Step 2. Click the Insert tab on the Ribbon
2-4 Line Charts, Column Charts, and Bar Charts 35

Step 3. Click the Insert Scatter (X,Y) or Bubble Chart button in the Charts group
When the list of chart subtypes appears, click the Bubble button
Then edit the chart as outlined in Section 2-2.
We plot the TSA wait time along the horizontal axis and the parking rate along the vertical
axis, and vary the size of each bubble to represent the number of enplanements. We see
that airports with fewer passengers tend to have lower wait times than those with more
passengers. There seems to be less of a relationship between parking rate and number of
passengers. Airports with lower wait times do tend to have lower parking rates.

FIGURE 2.9 A Bubble Chart using the Data in File AirportData

Cheapest Parking Rate (per day)

$25.00

$20.00

$15.00

$10.00

$5.00

$0.00
0.00 2.00 4.00 6.00 8.00 10.00 12.00
TSA Wait Time (minutes)

In a bubble chart, you might wish to change which variables correspond to the x (horizontal)
values, the y (vertical) values, and the bubble sizes. Once the chart has been created in Excel,
the following steps can be used to change these assignments.
Step 1. Right-click any bubble and choose Select Data...
Step 2. When the Select Data Source dialog box appears, click the Edit button under
Legend Entries (Series)
Step 3. Enter the location of the data you want to correspond to the horizontal values
in the Series X values: box (see Figure 2.10). Do not include column headers
Step 4. Repeat Step 3 for Series Y Values: box and the Series bubble size: box
Click OK

2-4 Line Charts, Column Charts, and Bar Charts

In this section, we consider the line chart, a natural extension of a scatter chart discussed in
the previous section. We also introduce column and bar charts which are useful for display-
ing categorical data.

Line Charts
A line chart uses a point to represent a pair of quantitative variable values, one
value along the horizontal axis and the other on the vertical axis, with a line connecting the
points. Line charts are very useful for time series data (data collected over a period of time:
minutes, hours, days, years, etc.). As an example, let us consider Cheetah Sports. Cheetah
sells running shoes and has retail stores in shopping malls throughout the United States. The
36 Chapter 2 Selecting a Chart Type

FIGURE 2.10 The Edit Series Dialog Box

file Cheetah contains the last ten years of sales for Cheetah Sports, measured in millions of
dollars. These data are shown in Figure 2.11. Figure 2.12 displays a scatter chart and a line
chart created in Excel for these sales data.

FIGURE 2.11 Data in the file Cheetah

Cheetah
 2-4 Line Charts, Column Charts, and Bar Charts 37

The following steps create the line chart of the Cheetah Sports sales data shown in
Figure 2.12b.
Step 1. Select cells A1:B11
Step 2. Click the Insert tab on the Ribbon
Step 3. In the Charts group, click the Insert Scatter (X,Y) or Bubble Chart
button
Select Scatter with Straight Lines and Markers
Edit the chart as described in Section 2-2
Comparing Figure 2.12b with Figure 2.12a, the addition of lines between the points sug-
gests continuity and makes it is easier for the reader to see and interpret changes that have
occurred over time.

FIGURE 2.12 Scatter Chart (a) and Line Chart (b) for Cheetah Sports Sales

Sales ($ millions)
250

200

150

100

50

0
0 2 4 6 8 10 12
Year
(a)

Sales ($ millions)
250

200

150

100

50

0
0 2 4 6 8 10 12
Year
(b)
38 Chapter 2 Selecting a Chart Type

Let us consider a second example that illustrates multiple lines on a single chart. Consider
the file CheetahRegion. Cheetah Sports has two sales regions: the eastern region and the
western region. The file breaks down the total sales for the ten-year period by region as
CheetahRegion shown in Figure 2.13.

FIGURE 2.13 The Data in the file CheetahRegion

Cheetah Sports sales by region are shown in Figure 2.14. To create the line chart shown in
Figure 2.14, select cells A1:C11 (do not select D1:D11) in the file CheetahRegion and follow
the Steps 2 and 3 previously outlined for constructing a line chart. In addition to the chart editing
from Section 2-2, we have also changed the color scheme to Monochromatic Palette 1 (using
the Chart Styles option as described in the Notes + Comments at the end of Section 2-2).
We can see from Figure 2.14 that sales in the Western region have increased over the
last three years of this ten-year period whereas sales in the Eastern region have dropped
substantially since year seven.

FIGURE 2.14 Line Chart for Cheetah Sports Sales by Region

Sales ($ millions) Eastern Sales Western Sales

140

120

100

80

60

40

20

0
0 2 4 6 8 10 12
Year
 2-4 Line Charts, Column Charts, and Bar Charts 39

Column Charts
A column chart displays a quantitative variable by category or time period using vertical
bars to display the magnitude of a quantitative variable. We have seen an example of a col-
umn chart in the zoo attendance data, where the categories are months of the year and the
quantitative variable is zoo attendance. Let us elaborate more about when to use a column
Cheetah
chart by continuing the Cheetah Sports annual sales example shown in Figure 2.11.
The following steps create the column chart of the Cheetah Sports sales data shown in
Figure 2.15.
Step 1. Select cells A1:B11
Step 2. Click the Insert tab on the Ribbon
Step 3. In the Charts group, click the Insert Column or Bar Chart button
Select Clustered Column
Excel displays the year as if it is a quantitative variable. To correct this, we need the
following steps:
Step 4. Right click the chart and select Change Chart Type…
Step 5. When the Change Chart Type task pane appears, select the Cluster Column
type that plots the appropriate number of variables (in this case, the single
variable Sales plotted with ten monochromatic columns) and click OK
Edit the chart as outlined in Section 2-2
The next step adds data labels to the bars.
Step 6. Click the Chart Elements button and select Data Labels

FIGURE 2.15 A Column Chart for Cheetah Sports Sales

Total Sales ($ millions)

195
183
177 175
170
154
145

110
87 90

1 2 3 4 5 6 7 8 9 10
Year

The line chart in Figure 2.12b, and the column chart in Figure 2.15, are both good displays
of the Cheetah Sports annual sales. The line chart, with its connected lines, makes it easier
to see how the sales are changing over time. The column chart, with its data labels, is pre-
ferred if it is important for the audience to know the values of sales in each year. Moreover,
adding data labels to a line chart generally makes the chart too cluttered. On the other
hand, if there are numerous categories or time periods, the line chart (without data labels)
would be preferred over the column chart with data labels because the column chart would
appear too cluttered and labels would not be readable.
40 Chapter 2 Selecting a Chart Type

Let us now reconsider the regional data for Cheetah Sports in the file CheetahRegion.
Using these data, let us construct a clustered column chart and compare it to the line chart
in Figure 2.14. A clustered column chart displays multiple quantitative variables by cat-
CheetahRegion
egories or time periods with different colors, with the height of the columns denoting the
magnitude of the quantitative variable.
To create the clustered column chart in Figure 2.16, select cells A1:C11 in the file
CheetahRegion as shown in Figure 2.13 (do not select cells D1:D11). Follow Steps 2–5
Clustered column charts with
previously outlined for a column chart. In addition to the chart editing from Section 2-2,
multiple variables are also
called side-by-side column we have also changed the color scheme to Monochromatic Palette 1 (using the Chart
charts. Styles option as described in the Notes + Comments at the end of Section 2-2).

FIGURE 2.16 Clustered Column Chart for Cheetah Sports Sales by Region

Sales ($ millions)
Eastern Sales Western Sales
140

120

100

80

60

40

20

0
1 2 3 4 5 6 7 8 9 10
Year

Comparing Figures 2.14 and 2.16, we see that the changes in sales within a region over
time are more apparent in the line chart. The clustered column chart in Figure 2.16 appears
cluttered and the changes in sales are not as obvious as in Figure 2.14. Adding data labels
to Figure 2.16 would make the clustered column chart even more cluttered.
While Figure 2.14 is preferred over Figure 2.16 for the regional sales data for Cheetah
Sports, neither of these charts convey that the Eastern and Western regions make up the
total sales. It is difficult to tell how total sales are changing. We make this more obvious by
using a stacked column chart. A stacked column chart is a column chart that uses color to
denote the contribution of each subcategory to the total.
To create a stacked column chart for Cheetah Sports, we select cells A1:C11 in the file
CheetahRegion, and repeat Steps 2–5 previously outlined for a column chart—except in
Step 3, we click the Insert Column or Bar Chart in the Charts group and select Stacked
Column . After chart editing, we obtain the stacked column chart shown in Figure 2.17.
This chart shows the combination of Eastern and Western region sales by year and the total
height of the column indicates the level of total sales.
The Cheetah Sports example with regional sales data demonstrates the important princi-
ple that the appropriate chart depends not only on the type of data, but also the goal of the
analysis and needs of the audience. If demonstrating the change in sales over time within
each region is a key point, then a line chart in this case is a good choice. If representing the
total sales level and how each region contributes to total sales over time is important, then a
stacked column chart is a good choice.
 2-4 Line Charts, Column Charts, and Bar Charts 41

FIGURE 2.17 Stacked Column Chart for Cheetah Sports Sales by Region

Sales ($ millions)
250
Eastern Sales Western Sales

200

150

100

50

0
1 2 3 4 5 6 7 8 9 10
Year

Bar Charts
A bar chart shows a summary of categorical data using the length of horizontal bars to
display the magnitude of a quantitative variable. That is, a bar chart is a column chart
turned on its side. Like column charts, bar charts are useful for comparing categorical
variables and are most effective when you do not have too many categories. Figure 2.2
in the Data Visualization Makeover of the Allocation of Funds in New York City is a
good example. As shown in that example, a bar chart can be a good substitute for a pie
chart when showing composition. Sorting the data as in Figure 2.2 makes the rank order
of the components by the magnitude of the quantitative variable more obvious. A bar
chart is preferred over a column chart if there are lengthy category names because it is
easier to display the names horizontally (for improved legibility). However, for time
series data, a column chart is better as it is more natural to display the passage of time
from left to right horizontally.
A clustered bar chart displays multiple quantitative variables for categories or time
periods using the length of horizontal bars to denote the magnitude of the quantitative
variables and separate bars and colors to denote the different variables. Like a stacked
column chart, a stacked bar chart is a bar chart that uses color to denote the contribution
of each subcategory to the total. As with column charts, clustered and stacked bar charts
are available in Excel by clicking on the Insert Column or Bar button in the Charts
group and then selecting either Clustered Bar or Stacked Bar .

N otes 1 C omments

1. In this section, we have shown how to use the Insert Scat- Sports data, which are numbers rather than actual years,
ter (X,Y) or Bubble Chart button and the Scatter with show up as a line on the chart rather than being interpreted
Straight Lines and Markers to construct a line chart. as the categories for the horizontal axis.
Another alternative is to use the Insert Line or Area 2. In Chapter 3, we discuss the issue of trying to present too
Chart button . This works for time series data (dates, much information on a single chart. In some cases, it may
months, years), but the option assumes numerical data is be preferable to use two similar charts rather than a stacked
to be graphed. For example, the periods in the Cheetah bar/column or clustered bar/column chart.
42 Chapter 2 Selecting a Chart Type

2-5 Maps
In this section, we introduce three types of maps used to display various types of data. You
are most likely familiar with geographic maps which are very useful for displaying data
that have a spatial or geographic component. We will also discuss heat maps and treemaps.
Each is available in Excel.

Geographic Maps
A geographic map is generally defined as a chart that shows characteristics and the
arrangement of the geography of our physical reality. A geographic map of the United
States shows state borders and how the states are arranged. A choropleth map is a geo-
graphic map that uses shades of a color, unique colors, or symbols to indicate quantitative
or categorical variables by geographic region or area.
Let us consider creating a choropleth map of the United States for which color shading
is used to denote the population of each state. A darker shade will indicate a higher popula-
tion and a lighter shade will indicate a lower population.
The population data for the fifty states can be found in the file StatePopulation. A por-
tion of the data set is shown in Figure 2.18. The fifty states in the United States are listed in
column A and the corresponding estimated population of each state is in column B.

FIGURE 2.18 A Portion of the State Population Data

StatePopulation

The following steps will create a choropleth map using shading to denote size of the popu-
lation for each state.
Step 1. Select cells A1:B51
Step 2. Click the Insert tab on the Ribbon
Step 3. In the Charts group, click the Maps button Maps

After selecting Filled Map, you Select Filled Map

may need to allow Excel to
send your request to the Bing
After editing the chart title as outlined in Section 2-2, we obtain the map in Figure 2.19. The
search engine to complete the map shows the states with the highest populations are California, Texas, Florida, and New York.
appropriate map. As an example of a choropleth map with categorical data, let us consider how Amazon
is able to deliver packages so quickly to its U.S. customers. Amazon distributes customer
orders from fulfillment centers which stock most of the products Amazon sells. The file
AmazonFulfill contains categorical data by state. For each state, “Yes” or “No” is provided
to denote whether or not that state has at least one Amazon fulfillment center. A portion of
the Amazon fulfillment data set is shown in Figure 2.20.
 2-5 Maps 43

FIGURE 2.19 A Choropleth Map Using Shading for State Population

Estimated United States Population by State

Estimated 2020 Population
39,937,489

567,025

Powered by Bing
© GeoNames, Microsoft, TomTom

FIGURE 2.20 A Portion of the Amazon Fulfillment Data

AmazonFulfill

Selecting cells A1:B51 and following Steps 2–3 listed in the state population example
results in the map shown in Figure 2.21. Amazon has at least one fulfillment center in 38
of the 50 states. The states without a fulfillment center tend to be either relatively sparsely
populated or a geographic outlier. Clearly, Amazon has a lot of fulfillment centers to ensure
quick customer delivery times for many of the products it sells.
Next we consider two useful types of maps that are not required to be geographic.
44 Chapter 2 Selecting a Chart Type

FIGURE 2.21 A Choropleth Map Using Color for a Categorical Variable

Which States Have At Least One Amazon Fullfillment Center?

Yes
No

Powered by Bing
© GeoNames, Microsoft, TomTom

Heat Maps
A heat map is a two-dimensional (2D) graphical representation of data that uses different
shades of color to indicate magnitude. Let us consider the data in file SameStoreSales. The
SameStoreSales data are shown in Figure 2.22. The rows of this data set correspond to store locations and

FIGURE 2.22 Same-Store-Sales Data in the File SameStoreSales

 2-5 Maps 45

the columns are the months of the year. The percentages given indicate the change in sales
over the same month last year for a given store. This percentage change metric is com-
monly used in the retail industry and is referred to as “same-store-sales.” For example, the
St. Louis store’s sales for January are 2% lower than last year in January.
Figure 2.23 shows a heat map of the same-store-sales data given in Figure 2.22. The
cells shaded red in Figure 2.23 indicate declining same-store sales for the month, and cells
shaded blue indicate increasing same-store sales for the month. The following steps create
the heat map shown in Figure 2.23.
Step 1. Select cells B2:M17
Step 2. Click the Home tab on the Ribbon.
Step 3. Click Conditional Formatting in the Styles group
Select Color Scales and click Blue-White-Red Color Scale

FIGURE 2.23 A Heat Map for Same-Store Sales

The heat map in Figure 2.23 helps the reader to easily identify trends and patterns. We can see
that Austin has had positive increases throughout the year, while Pittsburgh has had consistently
negative same-store sales results. Same-store sales at Cincinnati started the year negative but
then became increasingly positive after May. In addition, we can differentiate between strong
positive increases in Austin and less substantial positive increases in Chicago by means of color
shadings. A sales manager could use the heat map in Figure 2.23 to identify stores that may
require countermeasures and other stores that may provide ideas for best practices. Heat maps
can be used effectively to convey data over different areas, across time, or both, as seen here.

Treemaps
A treemap is a chart that uses the size, color, and arrangement of rectangles to display the
magnitudes of a quantitative variable for different categories, each of which are further
decomposed into subcategories. The size of each rectangle represents the magnitude of the
quantitative variable within a category/subcategory. The color of the rectangle represents
the category and all subcategories of a category are arranged together.
Categorical data that is further decomposed into subcategories is called hierarchical
data. Hierarchical data can be represented with a tree-like structure, where the branches
of the tree lead to categories and subcategories. As an example, let us consider the top ten
46 Chapter 2 Selecting a Chart Type

brand values given in the file BrandValues (source: Forbes.com). The data appear in the
file as shown in Figure 2.24. Each observation consists of an industry, a brand within an
industry, and the value of the brand.

FIGURE 2.24 Data in the file BrandValues

BrandValues

Figure 2.25 shows how these data have a hierarchical or tree structure. The base of the
tree is the top ten brand values. The category is the industry of each company, the subcate-
gory is the brand name, and the value of the brand is the quantitative variable.

FIGURE 2.25 The Hierarchical Tree Structure of the Top Ten Brand Values Data

Top 10 Brand Values

Industry Technology Beverages Leisure Automotive Restaurant

Company Apple Google Microsoft Amazon Facebook Samsung Coca-Cola Disney Toyota McDonald’s
Value ($ Billions) 205.5 167.5 125.3 97 88.9 53.1 59.2 52.2 44.6 43.8

Figure 2.26 is an example of a treemap for the brand values data. The following steps are
used to create a treemap in Excel using the data in the file BrandValues.
Step 1. Select cells A1:C11
Step 2. Sort the data by Industry by using the following steps:
Click Data on the Ribbon
Click the Sort button in the Sort & Filter group
In the Sort dialog box, select Industry from the drop-down menu
From the Order drop-down menu select A to Z
 2-6 When to Use Tables 47

Step 3. Click Insert on the Ribbon.

Click the Insert Hierarchy Chart button in the Charts group
Select Treemap from the drop-down menu
To display the brand values:
Step 4. Click any brand label and then right click
Select Format Data Labels… from the drop-down menu
Select Text Options, Label Options and select Value
The colors in the treemap shown in Figure 2.26 correspond to industries. Each of the rect-
angles represents a brand and the size of the rectangles indicates the size of the brand’s
value. We see that the technology industry has six brands in the top ten. Apple, Google, and
Microsoft are the three highest brand values.

FIGURE 2.26 A Treemap of the Ten Most Valuable Brands

The Ten Most Valuable Brands ($ Billions)

Automotive Beverages Leisure Restaurants Technology

Technology Beverages

Coca-Cola, 59.2
Leisure
Amazon, 97.0

Google, 167.7 Disney, 52.2

Automotive Restaurants

Facebook, 88.9

McDonald's,
Apple, 205.5 Microsoft, 125.3 Samsung, 53.1 Toyota, 44.6 43.8

2-6 When to Use Tables

Tables versus Charts
In general, charts can often convey information faster and easier to readers than tables, but
in some cases a table is more appropriate. Tables should be used when the:
●● reader needs to refer to specific numerical values.
●● reader needs to make precise comparisons between different values and not just rela-
tive comparisons.
●● values being displayed have different units or very different magnitudes.

Let us consider the case of Gossamer Industries. When the accounting department of Gos-
samer Industries is summarizing the company’s annual data for completion of its federal tax
forms, the specific numbers corresponding to revenues and expenses are important and not just
the relative values. Therefore, these data should be presented in a table similar to Table 2.1.
Similarly, if it is important to know by exactly how much revenues exceed expenses
each month, then this would also be better presented as a table rather than as a line chart as
seen in Figure 2.27. Notice that it is very difficult to determine the monthly revenues and
48 Chapter 2 Selecting a Chart Type

Table design is discussed in costs in Figure 2.27. We could add these values using data labels, but they would clutter the
Chapter 3. figure. A preferred solution is to combine the chart with the table into a single figure, as in
Figure 2.28, to allow the reader to easily see the monthly changes in revenues and costs
while also being able to refer to the exact numerical values.

TABLE 2.1  able Showing Exact Values for Costs and Revenues by Month
T
for Gossamer Industries
Month
Jan Feb Mar Apr May June Total
Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 326,567
Revenues ($) 64,124 66,125 67,125 48,178 51,785 55,678 353,015

Gossamer

FIGURE 2.27 A Line Chart of Monthly Costs and Revenues for Gossamer Industries

Gossamer Industries: Revenue and Costs

80,000 Costs ($) Revenues ($)

70,000

60,000

50,000

40,000

30,000

20,000

10,000

0
Jan Feb Mar Apr May June

We construct the chart-table Using the data in the file Gossamer, the following steps show how to create the line chart
combination using a line accompanied with a table as shown in Figure 2.28.
chart because this option
is not available for a scatter Step 1. Select cells A2:G4
chart. Excel does not support Step 2. Click the Insert tab on the Ribbon
integration of charts and
Step 3. Click the Insert Line or Area Chart button in the Charts group
tables for all chart types.
Step 4. When the list of column and bar charts subtypes appears, click the
Line button
Step 5. Click anywhere on the chart
Click the Chart Elements button
Select the check box for Data Table
Edit the chart as outlined in Section 2-2
 2-7 Other Specialized Charts 49

FIGURE 2.28 Combined Table and Line Chart of Monthly Costs and Revenues for Gossamer
Industries

Gossamer Industries: Revenue and Costs

80,000

70,000

60,000

50,000

40,000

30,000

20,000

10,000

0
Jan Feb Mar Apr May June
Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985
Revenues ($) 64,124 66,125 67,125 48,178 51,785 55,678

Displaying values with Now suppose that we wish to display data on revenues, costs, and head count for each
different units on the same
month. Costs and revenues are measured in dollars, but head count is measured in number
line chart is known as a dual-
axis chart. We will discuss
of employees. Although all of these values can be displayed on a line chart using multiple
these again in Chapter 9. vertical axes, this is generally not recommended. Because the values have widely different
magnitudes (costs and revenues are in the tens of thousands, whereas head count is approx-
imately 10 each month), it would be difficult to interpret changes on a single chart.
Therefore, a table similar to Table 2.2 is recommended.

TABLE 2.2  able Displaying Head Count, Costs, and Revenues, for
T
Gossamer Industries
Month
Jan Feb Mar Apr May June Total
Head Count 8 9 10 9 9 9
Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 326,567
Revenues ($) 64,124 66,125 67,125 48,178 51,785 55,678 353,015

2-7 Other Specialized Charts

In this section, we discuss three additional chart types: the waterfall chart, the stock chart,
and the funnel chart. Waterfall charts and stock charts are used primarily in financial ana-
lytics, while funnel charts are prominent in marketing and sales.

Waterfall Charts
A waterfall chart is a visual display that shows the cumulative effect of positive and nega-
tive changes on a variable of interest. The changes in a variable of interest are reported for a
series of categories (such as time periods) and the magnitude of each change is represented
by a column anchored at the cumulative height of the changes in the preceding categories.
50 Chapter 2 Selecting a Chart Type

FIGURE 2.29 Gossamer Data on Costs, Revenues, and Gross Profit

GossamerGP

Continuing with the Gossamer Industries example from the Section 2-6, consider the
data in the file GossamerGP. The data are shown in Figure 2.29. Gross profit is the differ-
ence between revenue and variable costs.
The following steps are used to create the waterfall chart of gross profit shown in Figure 2.30.
Step 1. Select cells A2:H2. Hold down the control key (Ctrl) and also select cells A5:H5
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Waterfall, Funnel, Stock, Surface or Radar Chart button
in the Charts group
When the list of subtypes appears, click the Waterfall button
In the initial chart, notice that the Total has been treated like another month. The following
steps will make the total appear as in Figure 2.30.
Step 4. Double-click the column Total to activate the Format Data Series task pane, and
then click the column of data again to activate the Format Data Point task pane
Step 5. When the Format Data Point task pane appears, click the Series Options
button
Select the check box for Set as total
Then edit the chart as outlined in Section 2-2

FIGURE 2.30 A Waterfall Chart for the Gossamer Gross Profit Data

Gossamer Gross Profit ($)

Increase Decrease Total
3,000

9,667 4,693 26,448

–3,980

–2,933

16,001

Jan Feb Mar Apr May June Total

 2-7 Other Specialized Charts 51

Figure 2.30 shows the gross profit by month, with blue indicating a positive gross profit and
orange indicating a negative gross profit. The upper or lower level of the bar indicates the
cumulative level of gross profit. For positive changes, the upper level of the bar is the cumula-
tive level, and for negative changes, the lower end of the bar is the cumulative level. Here we
see that cumulative level of gross profit rises from January to March, drops in April and May
and then increases in June to the cumulative gross profit of $26,448 for the six-month period.

Stock Charts
A stock chart is a graphical display of stock prices over time. Let us consider the stock price
data for telecommunication company Verizon Communications given in the file Verizon. As
shown in Figure 2.31, this data set lists, for five trading days in April: the date, opening price
per share (price per share at the beginning of the trading day), the high price (highest price per
share observed during the trading day), the low price (the lowest price per share observed dur-
ing the trading day), and the closing price (the price per share at the end of the trading day).

FIGURE 2.31 Stock Price Data for Verizon Communications

Verizon

Excel also provides an open- Excel provides four different types of stock charts. We illustrate the simplest one here, the
high-low-close stock chart, a high-low-close stock chart. A high-low-close stock chart is a chart that shows the high
volume-high-low-close stock
chart, and a volume-open-
value, low value, and closing value of the price of a share of a stock at several points in
high-low-close chart. These time. The difference between the highest and lowest share prices for each point in time is
charts add data on a stock’s represented by a vertical bar, and the closing share price by a marker on the bar.
opening price and trading The following steps are used to create Figure 2.32, the high-low-close stock chart for
volume to the basic high-low- the Verizon stock price data.
close stock chart.
Step 1. Select cells A1:A6. Hold down the control key (Ctrl) and also select cells C1:E6
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Waterfall, Funnel, Stock, Surface or Radar Chart button
in the Charts group
When the list of subtypes appears, click the High-Low-Close button
Edit the chart using steps outlined in Section 2-2
The following steps add the closing price labels and markers.
Step 4. Click the Chart Elements button and select Data Labels
Step 4 places three sets of labels on each vertical bar (highest, closing and lowest price per
share). The following steps clean up the display.
Step 5. Click any of the high price per share labels and press the Delete key. Do the
same for the low price per share labels
Step 6. On one of the vertical lines, click a data point directly next to one of the
closing price labels
52 Chapter 2 Selecting a Chart Type

Step 7. When the Format Data Series task pane appears, click the Fill &Line button
, then click Marker
Under Fill, select Solid fill and to the right of Color, select black from
the drop down menu
Under Border, select Solid line, and to the right of Color, select Black
While typically used to display from the drop down menu
stock price data over time,
To the right of Width, select 3 pt
a high-low-close stock chart
can also be used to display As shown in Figure 2.32, the closing prices per share over the five days are given. We see
the maximum, minimum, and
that on April 20, 21, and 23, the price closed near the low end of the trading price range.
mean (or median) of a variable
of interest measured over a
On April 24, the closing price was near the highest price of the day. On April 22, the clos-
set of categories. ing price was near the middle of the trading price range.

FIGURE 2.32 A High-Low-Close Stock Chart for Verizon Wireless

Price per Share ($)

59.50 Close

59.00

58.50

58.13
58.00 57.99 57.93
57.59
57.50

57.00
56.82
56.50

56.00

55.50
20-Apr 21-Apr 22-Apr 23-Apr 24-Apr
Date

Funnel Charts
Another specialized chart is a funnel chart. A funnel chart shows the progression of a
quantitative variable for various categories from larger to smaller values. A funnel chart
is often used to show the progression of sales leads that are converted through a series of
steps to an eventual sale, but any progression of larger values to smaller values over a series
of nested categories can be illustrated with a funnel chart. As an illustration, let us consider
a company whose goal is to grow the number of well-qualified members on its data science
team. The hiring process involves the following steps: (1) post the job ad and candidates
apply and are then referred to as applicants, (2) applicants are given a technical test; those
who pass are deemed technically qualified, (3) the technically qualified set of applicants
are invited to do Zoom interviews, and based on the Zoom interviews, a subset of the tech-
nically qualified applicants are deemed finalists and are invited for on-site interviews,
(4) based on test scores and the on-site interviews, a subset of the finalists are offered
employment, and (5) those who accept are hired.
The data for this process are in the file DataScienceSearch shown in Figure 2.33. A
funnel chart of these data is shown in Figure 2.34. First, we give the steps for creating this
chart and then we provide a brief summary of the chart.
 2-7 Other Specialized Charts 53

FIGURE 2.33 Data for Data Scientist Hiring

DataScienceSearch

The following steps are used to create the funnel chart shown in Figure 2.34.
Step 1. Select cells A1:B6.
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Waterfall, Funnel, Stock, Surface or Radar Chart button
in the Charts group
When the list of subtypes appears, click the Funnel Chart button
Edit the chart using steps outlined in Section 2-2.

FIGURE 2.34 A Funnel Chart for Data Scientist Hiring

Data Science Search

Applicants 51

Technically Qualified 37

Finalists 12

Offers 7

Hired 4

The funnel chart shows the narrowing of the field of applicants as the process progresses.
We see that the process started with 51 applicants and ended with 4 new hires. Specifically,
we observe that the Zoom interviews narrowed the field from 37 technically qualified
applicants to 12 finalists who were invited to interview on-site.
54 Chapter 2 Selecting a Chart Type

2-8 A Summary Guide to Chart Selection

In this section, we conclude the chapter with a discussion summarizing guidelines for chart
selection.

Guidelines for Selecting a Chart

Recognizing that there are often exceptions to rules and that there are often disagreements
even among data visualization experts, we provide general recommendations based on the
goal of the visualization and the type of data being analyzed.
Goal: To Show a Relationship
–5% –6%
16% 15%
–9% –6%

scatter bubble line stock column bar heat map

To show a relationship between two quantitative variables, we recommend a scatter chart. An

example is the temperature and snowfall data shown in a scatter chart in Figure 2.7. When
dealing with three quantitative variables, a bubble chart can be used. Line charts can be used to
emphasize the pattern across consecutive data points and are commonly used to display relation-
ships over time. Stock charts show the relationship between time and stock price. Column charts,
bar charts, and heat maps can be used to show the relationships that exist between categories.
Goal: To Show Distribution

scatter bubble column bar choropleth map

In addition to being useful for showing the relationships between quantitative variables, scat-
ter and bubble charts can be useful for showing how the quantitative variable values are dis-
tributed over the range for each variable. For example, from the scatter chart in Figure 2.7, we
can see that only 2 of the 51 cities have an average annual snowfall greater than 80 inches.
Other chart types useful Column and bar charts can be used to show the distribution of a variable of interest over
for showing how data are discrete categories or time periods. For example, Figure 2.5 shows the distribution of zoo
distributed that rely on more
attendance by time (month). As previously mentioned, column charts rather than bar charts
advanced statistical concepts
are discussed in Chapter 5.
should be used for distribution over time, as it is more natural represent the progression of
time from left to right. A choropleth map shows the distribution of a quantitative or cate-
gorical viable over a geographic space. Figures 2.19 and 2.21 are examples of these.
Goal: To Show Composition

bar stacked bar stacked column treemap waterfall funnel

When the goal is to show the composition of an entity, a good choice is a bar chart, sorted by
contribution to the whole. An example is the New York City budget in Figure 2.2. A stacked
bar chart is appropriate for showing the composition of different categories and a stacked
column chart is good for showing composition over a time series. Figure 2.17, the sales for
Cheetah Sports by region is a good example of a stacked column chart with time series data.
While the goal of a pie chart A treemap shows composition in the situation where there is a hierarchical structure
is to show composition, for among categorical variables. In Figure 2.26, we see the brand values (the quantitative vari-
reasons discussed in the able of interest) for companies within industry sectors. For example, the technology sector is
next section, we do not composed of six brands in the top ten. All other sectors are composed of only a single brand.
recommend the use of pie
charts.
A waterfall chart shows the composition of a quantitative variable of interest over time
or category. For example, Figure 2.30 shows the composition of the final value of gross
profit over time. A funnel chart also shows composition in the sense that going from the
bottom of the funnel to the top gives the composition of the original set at the top of the
funnel. The funnel chart for the hiring process in Figure 2.34 is an example.
 2-8 A Summary Guide to Chart Selection 55

Goal: To Show Ranking

bar column
Bar charts and column charts, sorted on the cross-sectional quantitative data of interest
across categories, can be used to effectively show the rank order of categories on the quan-
titative variable. An example is the ten categories ranked by spending allocation in the
New York City budget shown in Figure 2.2.
When trying to select a chart type, we recommend starting with understanding the needs
of the audience to determine the goal of the chart, understanding the types of data you
have, and then selecting a chart based on the guidance provided in this section. Like most
analytics tools, it is important to experiment with different approaches before arriving at a
final decision on your data visualization.

Some Charts to Avoid

In this section, we discuss some charts that should be avoided. There are charts that many
data visualization experts agree should be avoided. Usually this is because a chart is overly
cluttered or takes too much effort for most audiences to interpret the chart quickly and
accurately. Here we provide some guidance on charts we believe should be avoided in
favor of other types of charts.
As we have already discussed in the data visualization makeover at the beginning of
this chapter, many experts suggest that pie charts should be avoided. Instead of a pie chart,
consider using a bar chart. This is because science has shown that we are better at assessing
differences in length than angle and area. Small differences can be better detected in length
than area, especially when sorted by length. Also, using a bar chart simplifies the chart in
that there is no longer a need for a different color for each category. Figures 2.1 and 2.2
show the difference between the pie chart and the bar chart and illustrate why the latter is
preferred.
A radar chart is also referred Another chart to be avoided is a radar chart. A radar chart is a chart that displays mul-
to as a spider chart or a web
tiple quantitative variables on a polar grid with an axis for each variable. The quantitative
chart.
values on each axis are connected with lines for a given category. Multiple categories can
be overlaid on the same radar chart.
Let us consider data on four suppliers of a component needed by Newton Industries.
Newton manufactures high-performance desktop computers and has started to vet four pos-
sible suppliers of one of the components needed for its computers. Newton’s management
needs to select a supplier to provide the component and has collected data on the percent-
age of late shipments, the percentage of defective components delivered and the cost per
unit each supplier would charge. These data are in the file NewtonSuppliers and are shown
in Figure 2.35. Figure 2.36 is the radar chart created from these data.

FIGURE 2.35 Supplier Performance Data for a Component for Newton

Industries

NewtonSuppliers
56 Chapter 2 Selecting a Chart Type

The radar chart in Figure 2.36 has three axes corresponding to the three columns of data in
Figure 2.35. Luckily the three variables are of roughly the same magnitude. Variables of very
different scales can distort a radar chart. The four suppliers each have their own color and
their data are connected by lines. Since Newton presumably wants low values for percentage
late, percentage of defective components and cost per unit, a dominant supplier’s rectangle
would be completely inside its competitors. It appears from Figure 2.36 that the supplier
Foster might be the best choice, but it is difficult to distinguish the cost per unit. Even with
this very small data set, the radar chart is quite busy and difficult for an audience to interpret.

FIGURE 2.36 A Radar Chart of Supplier Performance for a Component

for Newton Industries

Supplier Performance
Ace Beaty Foster Rolf
% Late
12

10

Cost per unit ($) % Defective

Perhaps a better choice is the clustered column chart shown in Figure 2.37. Here we can
see that Foster is clearly better on percentage late and percentage defective and competitive
on price. We do note that for more suppliers, even the clustered column chart will become

FIGURE 2.37 A Clustered Column Chart of Supplier Performance for a Component for Newton
Industries

Supplier Performance
% Late % Defective Cost per Unit ($)
12

10

0
Ace Beaty Foster Rolf
Supplier
 2-8 A Summary Guide to Chart Selection 57

cluttered. Not surprisingly, manufacturers will often develop a scoring model so that a sin-
gle score can be computed and used to compare suppliers.
In Chapter 3, we will provide In addition to too much clutter and problems with scaling, another criticism of radar
a more detailed discussion of charts is that as the number of factors increases, they become more circular and suffer
how to remove clutter from
charts.
from the same criticisms as pie charts. Finally, although not as obvious in our three-
factor example, the order of the axes in a radar chart can dramatically alter the picture
presented by a radar chart and hence the audience’s perception. For these reasons, we
suggest avoiding the use of radar charts.
Another chart that many find difficult to read is an area chart. An area chart is a line
chart with the area between the lines filled with color. Figure 2.38 is an area chart of the
Cheetah Regional sales data shown in Figure 2.13. Area charts display volume and convey
continuity, but a simpler line chart such as Figure 2.14 or a stacked column chart such as
Figure 2.17 provide less cluttered alternatives.

FIGURE 2.38 An Area Chart for the Cheetah Sports Regional Sales Data

Sales ($ millions)
250
Eastern Sales Western Sales

200

150

100

50

0
1 2 3 4 5 6 7 8 9 10
Year

Dual-axis charts are one form Excel also provides combination charts called combo charts. A combo chart com-
of a combo chart; these are bines two separate charts, for example, a column chart and a line chart, on the same chart.
discussed in Chapter 9.
Combo charts can be overly cluttered and difficult to interpret, especially when they con-
tain both a left and right vertical axis.
Finally, we recommend always avoiding unnecessary dimensionality on any of the
Unnecessary use of
dimensionality and other chart
charts you select. Many Excel charts come in 2-dimensional (2D) and 3-dimensional (3D)
design issues are discussed in versions. We recommend avoiding 3D versions as the third dimension typically adds no
more detail in Chapter 3. additional understanding and can lead to more clutter.

Excel’s Recommended Charts Tool

Excel provides guidance for chart selection through its Recommended Charts tool. The
Recommended Charts button is found in the Charts group of the Insert tab on the
Ribbon. The following steps demonstrate the use of the Recommended Charts tool using
Zoo the zoo attendance data in the file Zoo shown in Figure 2.3.
58 Chapter 2 Selecting a Chart Type

Step 1. Select cells A1:B13

Step 2. Click the Insert tab on the Ribbon Recommended
Step 3. Click the Recommended Charts button Charts in the Charts group
The Insert Chart dialog box appears as shown in Figure 2.39. Four different chart types
are recommended, a column chart (also shown to the right), a bar chart, a funnel chart, and
a combination (combo) chart. Clicking on any of the four charts on the left will display that
chart enlarged and to the right, in the same way the column chart is displayed on the right in
Figure 2.39. This allows you to see an enlarged version of the chart before committing to the
chart.
Step 4. Select the Clustered Column chart and click OK
Edit the chart as outlined in Section 2-2

FIGURE 2.39 The Insert Chart Task Pane for the Zoo Attendance Data
 Summary 59

N otes 1 C omments

1. After clicking the Recommended Charts button, observe that Indeed, sometimes chart types that we would not recom-
selecting the All Charts tab from the Insert Chart task pane mend show up as choices under Recommended Charts.
generates a listing of all available chart types. Selecting any Two examples are apparent in the last two choices shown in
of the listed charts provides a preview of the selected chart. Figure 2.39. Using a funnel chart for the zoo attendance is
Hence an alternative to navigating the Charts group on the a poor choice, as there is no natural progression from high
Insert tab on the Ribbon is to click the Recommended Charts values to low values. Likewise, notice the combo chart sorts
button, select the All Charts tab and select from the list. the months by decreasing order of attendance, which is not
2. The Recommended Charts tool does not always recom- likely to be useful for these time series data if the goal is
mend chart types consistent with the advice in this chapter. to better understand the pattern of attendance over time.

S U M M A RY

In this chapter, we discussed how the goal of the analysis and the type of data should inform
chart selection. We provided detailed steps to create and edit charts in Excel. We discussed
a variety of popular chart types and provided steps for creating these charts in Excel.
A scatter chart displays pairs of quantitative variables and is very useful for detecting
patterns. A bubble chart is a scatter chart that represents a third quantitative variable by dif-
ferent size dots, known as bubbles. A line chart is a scatter chart with lines connecting the
points. A line chart, like a scatter chart, is good for detecting patterns and is very useful for
time series data. Line charts, by connecting the dots representing data points, provide more
of a sense of continuity than scatter charts.
A column chart displays a quantitative variable by using the height of the column to
denote the magnitude of the quantitative variable by category or time period. A clustered
column chart is a column chart that displays multiple quantitative variables using different
colors and side-by-side columns. A stacked column chart is a column chart that shows com-
position for each column by using color to denote subcategory contributions to the total.
Similar to a column chart, a bar chart displays the magnitude of a quantitative variable
using length, but by using horizontal bars rather than vertical columns. A clustered bar chart
and a stacked bar chart are similar to their column-chart counterparts, but they use horizon-
tal bars rather than vertical columns to denote the magnitude of the quantitative variable.
We also discussed three types of maps. A choropleth map is a geographic map that uses
shades of a color, different colors, or symbols to indicate quantitative or categorical vari-
ables by geographic region or area. A heat map is a two-dimensional graphical represen-
tation of data that uses different shades of color to indicate magnitude. Finally, a treemap
uses different-sized rectangles and color to display quantitative data that is associated with
hierarchical categories. We briefly discussed when to use a table or a combination of a
table and a chart, rather than a chart. If exact values are needed, a table or table/chart com-
bination might be the best choice.
Three specialized charts, waterfall, stock, and funnel were discussed. A waterfall chart
shows the cumulative effect of positive and negative changes on a variable of interest. A
stock chart displays various information about the share price of a stock over time. For
example, a high-low-close stock chart shows the high value, low value, and closing value
of the price of a share of stock over time. A funnel chart shows the progression of a quanti-
tative variable across various nested categories from larger to smaller values.
We provided guidance on how to select an appropriate chart based on the goal of the
chart and the type of data being displayed. We also discussed some chart types to avoid.
We concluded the chapter with a discussion of the Recommended Charts tool in Excel.
60 Chapter 2 Selecting a Chart Type

G L O S S A RY

Area chart A line chart with the area between the lines filled with color.
Bar chart A chart that displays a quantitative variable by category using the length of
horizontal bars to display the magnitude of a quantitative variable.
Bubble chart A scatter chart that displays a third quantitative variable using different sized
dots, which we refer to as bubbles.
Choropleth map A geographic map that uses shades of a color, different colors, or
symbols to indicate quantitative or categorical variables by geographic region or area.
Clustered bar chart A chart that displays multiple quantitative variables for categories or
time periods using the length of horizontal bars to denote the magnitude of the quantitative
variables and separate bars and colors to denote the different categories.
Clustered column chart A chart that displays multiple quantitative variables for
categories or time periods with different colors, with the height of the columns denoting
the magnitude of the quantitative variable.
Column chart A chart that displays a quantitative variable by category or time period
using vertical bars to display the magnitude of a quantitative variable.
Combo chart A chart that combines two separate charts, for example, a column chart and
a line chart, on the same chart.
Funnel chart A chart that shows the progression of a quantitative variable for various
nested categories from larger to smaller values.
Geographic map A chart that shows characteristics and the arrangement of the geography
of our physical reality.
Heat map A two-dimensional graphical representation of data that uses different shades
of color to indicate magnitude.
Hierarchical data Data that can be represented with a tree-like structure, where the
branches of the tree lead to categories and subcategories.
High-low-close stock chart A chart that shows the high value, low value, and closing
value of the price of a share of stock over time.
Line chart A chart that uses a point to represent a pair of quantitative variable values, one
value along the horizontal axis and the other on the vertical axis, with a line connecting the
points.
Radar chart A chart that displays multiple quantitative variables on a polar grid with an
axis for each variable. The quantitative values on each axis are connected with lines for a
given category.
Scatter chart A graphical presentation of the relationship between two quantitative
variables. One variable is shown on the horizontal axis and the other is shown on the
vertical axis and a symbol is used to plot ordered pairs of the quantitative variable values.
Stacked bar chart A bar chart that uses color to denote the contribution of each
subcategory to the total.
Stacked column chart A column chart that shows part-to-whole comparisons, either
over time or across categories. Different colors, or shades of color, are used to denote the
different parts of the whole within a column.
Stock chart A graphical display of stock prices over time.
Treemap A chart that uses the size, color, and arrangement of rectangles to display the
magnitudes of a quantitative variable for different categories, each of which are further
decomposed into subcategories. The size of each rectangle represents the magnitude of the
quantitative variable within a category/subcategory. The color of the rectangle represents
the category and all subcategories of a category are arranged together.
Waterfall chart A visual display that shows the cumulative effect of positive and negative
changes on a variable of interest. The basis of the changes can be time or categories and
changes are represented by columns anchored at the previous time or category’s cumulative
level.
Problems 61

P R O B L E M S

Conceptual
1. Sales by Region. Consider the following data for percentage of sales by region. LO 3

Percentage of
Sales Region Total Sales
East 28%
North 14%
South 36%
West 22%

a. Should a bar chart or a pie chart be used to display these data? Explain.
b. List two ways to enhance the formatting of the chart to improve interpretability.
2. Academic Makeup of Departments. You are conducting an analysis of the makeup
of the departments in your firm. Your goal is to compare the departments’ mixes of
academic backgrounds. You have defined the following categories for academic back-
ground: Business, Engineering, and Other. You have the percentage of employees in
each category for each of the four departments as shown in the table below. LO 3

Department Business Engineering Other

A 84% 0% 16%
B 45% 43% 12%
C 48% 20% 32%
D 17% 68% 15%

What type of chart is best suited to display these data?

3. Charts of Gasoline Prices. The following charts both show the average price (in dol-
lars) per gallon of gasoline in the United States for 36 consecutive months. Consider
the following charts. The first is a line chart and the second is a column chart. LO 3
Average Price per Gallon ($)
4.50

4.00

3.50

3.00

2.50

2.00

1.50

1.00

0.50

0.00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
Month
62 Chapter 2 Selecting a Chart Type

Average Price per Gallon ($)

4.50

4.00

3.50

3.00

2.50

2.00

1.50

1.00

0.50

0.00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
Month

Which better displays the data? Why?

i. Line chart
ii. Column chart
4. Pickup Truck Sales. The following charts display the sales of pickup trucks in the
United States by manufacturer for one year (source: The Wall Street Journal). You
must choose one of these charts to edit to produce the final chart for your presentation
to management. LO 3, 5

Number of Vehicles Number of Vehicles

1,000,000
900,000
800,000
700,000
600,000
500,000
400,000
300,000
200,000
100,000
0
Ford F-Series GM Chevy Fiat Chrysler Ram Toyota Tundra Nissa Titan
Silverado/GMC
Sierra

Column Chart

Fiat Chrysler Ram Ford F-Series

GM Chevy Silverado/GMC Sierra Nissa Titan
Toyota Tundra
Pie Chart

Number of Vehicles Number of Vehicles

Ford F-Series
Fiat Chrysler Ram 633,694

GM Chevy
Silverado/GMC Sierra
Food F-Series 896,526
Fiat Chrysler Ram

GM Chevy Silverado/
Toyota Tundra 807,923
GMC Sierra

Nissa Titan
Nissa Titan
0 200,000 400,000 600,000 800,000 1,000,000

Bar Chart
Toyota Tundra

Funnel Chart

Which chart is best for displaying these data? Explain your answer.
i. Column chart
ii. Pie chart
iii. Bar chart
iv. Funnel chart
Problems 63

5. NCAA Women’s Basketball. Since 1994, the NCAA Division I women’s basketball tour-
nament has had a starting field of 64 teams and, over the course of 63 single-elimination
games, a champion is determined. The following two charts (a funnel chart and a bar chart)
show how the tournament progresses from the starting field of 64 teams. LO 3
NCAA Women’s Basketball Tournament
First Round 64

Second Round 32

Sweet 16 16

Elite Eight 8

Final Four 4

Final 2
Champion 1

Funnel Chart

NCAA Women’s Basketball Tournament

First Round 64
Second Round 32
Sweet 16 16
Elite Eight 8
Final Four 4
Final 2
Champion 1

Bar Chart

Which chart best conveys these data? Why?

i. Funnel chart
ii. Bar chart
6. Worldwide Robot Supply. The International Federation of Robotics estimates the
worldwide supply of industrial robots each year. The charts below show estimates of the
worldwide supply of industrial robots for the years 2009 to 2021. The data are shown in
two different charts below, first is a line chart and the second a column chart. LO 3
Number of Robots (1000s units)
700

600

500

400

300

200

100

0
2009 2011 2013 2015 2017 2019 2021
Year
64 Chapter 2 Selecting a Chart Type

Industrial Robot Supply (1000s units)

630

553
484
421
381

294
254
221
166 178
159
121
60

2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Year

Which chart best displays these data? Why?

i. Line chart
ii. Column chart
7. R&D Project Portfolio. The Ajax Company uses a portfolio approach to manage their
research and development (R&D) projects. Ajax wants to keep a mix of projects to bal-
ance the expected return and risk profiles of their R&D activities. Consider a situation
in which Ajax has six R&D projects as characterized in the following table.
Each project is given an expected rate of return and a risk assessment, which is a value
between 1 and 10, where 1 is the least risky and 10 is the riskiest. Ajax would like to
visualize their current R&D projects to keep track of the overall risk and return of their
R&D portfolio. LO 3

Expected Rate of Capital Invested

Project Return (%) Risk Estimated ($ millions)
1 12.6 6.8 6.4
2 14.8 6.2 45.8
3 9.2 4.2 9.2
4 6.1 6.2 17.2
5 21.4 8.2 34.2
6 7.5 3.2 14.8

Which of the following chart types would be the most appropriate for these data?
Explain your answer.
i. Stacked bar chart
ii. Line chart
iii. Bubble chart
iv. Funnel chart
8. E-Marketing Campaign. Gilbert Furniture has initiated a new marketing campaign
for its high-end desk lamp. The analyst for e-commerce, Lauren Stevens, has been
tracking the progress of the campaign and has collected the following data based on an
email sent to the customer list: 68% opened the email, 29% clicked on the web link in
the email, 11% added the desk lamp to their cart, and 9% purchased the lamp. LO 3
Which of the following is the most appropriate chart for these data?
i. Funnel chart
ii. Stacked bar chart
 Problems 65

iii. Line chart

iv. Bubble chart
9. Choosing the Best Chart Type. Choose the most appropriate chart type (bar chart,
bubble chart, choropleth map, line chart) for each data set described below. Use each
chart type exactly once. LO 3
a. Advertising budget, number of salespeople, percent of market share for 10 products
b. Annual demand for potato chips (in tons) in the United States by state
c. Annual sales in millions of dollars for seven regional salespeople
d. Population of the United States for each year 1900–2020
10. Disney Ticket Prices. The three charts below show the price of a general admission
ticket to Walt Disney World for the years 2000 to 2020. The first is a bar chart, the
second is an area chart, and the third is a scatter chart. LO 3
Year
2020
2019
2018
2017
2016
2015
2014
2013
2012
2011
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
0 20 40 60 80 100 120 140
Disney Ticket Price ($)

Disney Ticket Price ($)

140

120

100

80

60

40

20

0
2000 2005 2010 2015 2020
Year
66 Chapter 2 Selecting a Chart Type

Disney Ticket Price ($)

140

120

100

80

60

40

20

0
2000 2005 2010 2015 2020
Year

Which type of chart best displays these data? Why?

i. Bar chart
ii. Area chart
iii. Scatter chart
11. Exploring Private Colleges. For a sample of 103 private colleges, data have been
collected on year founded, tuition and fees (not including room and board), and the
percentage of undergraduates who obtained their degree within six years (source: The
World Almanac). The following two charts plot tuition versus year founded, and gradu-
ation rate versus year founded, respectively. LO 4
Tuition ($)
50,000

45,000

40,000

35,000

30,000

25,000

20,000

15,000

10,000

5,000

0
1600 1700 1800 1900 2000
Year Founded
Problems 67

Graduation Rate (%)

100

90

80

70

60

50

40

30

20

10

0
1500 1600 1700 1800 1900 2000
Year Founded

a. The two charts are the same type of chart. What type of chart are these?
i. Line chart
ii. Scatter chart
iii. Stock chart
iv. Waterfall chart
b. Which of the following statements best describes the relationship between tuition
and year founded?
i. Private colleges founded before 1800 are expensive, but there are greater differ-
ences in tuition for private colleges founded after 1800.
ii. There is no apparent relationship between year founded and tuition for private
colleges.
iii. The newer the private college, the higher the tuition.
c. Which of the following best describes the relationship between graduation rate and
year founded?
i. There is no apparent relationship between year founded and graduation for private
colleges.
ii. The newer the private college, the higher the graduation rate.
iii. Private colleges founded before 1800 have high graduation rates, but there are
greater differences in graduation rate for private colleges founded after 1800.
12. Vehicle Production Data. The International Organization of Motor Vehicle Man-
ufacturers (officially known as the Organisation Internationale des Constructeurs
d’Automobiles, OICA) provides data on worldwide vehicle production by manufac-
turer. The following three charts, a line chart, a line chart with a table, and a clustered
column chart, show vehicle production numbers for four different manufacturers for
five recent years. LO 3
68 Chapter 2 Selecting a Chart Type

Number of Autos (millions) TOYOTA GM VOLKSWAGEN HYUNDAI

10
9
8
7
6
5
4
3
2
1
0
Year 1 Year 2 Year 3 Year 4 Year 5

Number of Autos (millions) TOYOTA GM VOLKSWAGEN HYUNDAI

10
9
8
7
6
5
4
3
2
1
0
Year 1 Year 2 Year 3 Year 4 Year 5
TOYOTA 8.04 8.53 9.24 7.23 8.56
GM 8.97 9.35 8.28 6.46 8.48
VOLKSWAGEN 5.68 6.27 6.44 6.07 7.34
HYUNDAI 2.51 2.62 2.78 4.65 5.76
 Problems 69

Number of Autos (millions)

10

0
Year 1 Year 2 Year 3 Year 4 Year 5
TOYOTA 8.04 8.53 9.24 7.23 8.56
GM 8.97 9.35 8.28 6.46 8.48
VOLKSWAGEN 5.68 6.27 6.44 6.07 7.34
HYUNDAI 2.51 2.62 2.78 4.65 5.76

Which type of chart best displays these data? Why?

i. Line chart
ii. Line chart with table
iii. Clustered column chart with table
13. Smartphone Ownership in Emerging Countries. Suppose we have the following
survey results regarding smartphone ownership by age in emerging countries. LO 3, 4

Other Cell No Call

Age Category Smartphone (%) Phone (%) Phone (%)
18–24 49 46 5
25–34 58 35 7
35–44 44 45 11
45–54 28 58 14
55–64 22 59 19
65+ 11 45 44
70 Chapter 2 Selecting a Chart Type

Cell Phone Ownership by Age Smartphone Other Cell No Cell Phone

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
18–24 25–34 35–44 45–54 55–64 65+
Age Category

Cell Phone Ownership by Age Smartphone Other Cell No Cell Phone

70%
60%
50%
40%
30%
20%
10%
0%
18–24 25–34 35–44 45–54 55–64 65+
Age Category

a. Which of the following charts best displays these data?

i. Stacked column
ii. Clustered column
b. Is the following statement true or false? Older people are less likely to own a smart-
phone than a different type of cell phone than younger people.
14. Home Goods Demand. In supply chain planning, demand is often measured in pounds
shipped. The following choropleth map shows the demand for home goods, measured
in millions of pounds shipped for each state in the United States. LO 4
a. Which five states have the highest demand?
b. Compare this map with Figure 2.19, which shows estimates of state population.
Does the demand for home goods appear to be related to population? Explain.
 Problems 71

Home Goods Demand Demand (in millions of pounds

shipped)
WA 87.6
MT ME
ND
OR MN
ID
SD WI NY 3.7
WY MI
IA PA
NV NE OH
UT IL IN
CA WV
CO KS VA
MO KY
NC
AZ TN
NM OK SC
AR
MS AL GA
TX
LA
FL

Powered by Bing
© GeoNames, Microsoft, TomTom

15. Coca-Cola Stock Prices. The following stock chart shows stock price performance for
Coca-Cola over a two-week period. Note that May 16 and May 17 are a Saturday and a
Sunday and are non-trading days. LO 4

Coca-Cola Stock Price Performance

Price per Share ($) Close
46.50

46.00
45.89
45.50 45.54
45.17
45.00 44.97 45.03
44.82
44.50 44.54

44.00 43.94
43.70
43.50
43.26
43.00

42.50
11-May 12-May 13-May 14-May 15-May 16-May 17-May 18-May 19-May 20-May 21-May 22-May
Date

a.
What type of chart is this?
b.
Which day seems to have the most intra-day price variability?
c.
What is the closing price on May 22?
d.
If you bought 100 shares at the closing price on May 19 and sold all of those shares
at the closing price on May 22, how much did you gain or lose (ignoring any trans-
action fees)?
16. Day Trading. In addition to the high, low, and closing price, an open-high-low-close
stock chart uses the opening price per share to give an indication of the net change in
the stock price from open to close on a given day. This is designated by a box inside the
72 Chapter 2 Selecting a Chart Type

high-low range. The range of the box is determined by the opening and closing price
per share. A black box indicates a loss and a white box indicates a gain for that day.
The length of the box indicates the magnitude of the loss or gain in share price. The
following chart is an open-high-low-close chart for a two-week period for Coca-Cola.
Note that May 16 and May 17 are a Saturday and a Sunday and are non-trading days.
Day trading is the practice of purchasing and then selling stock within the same day.
As a novice day trader, your strategy is to buy at the start of the day and sell at the end
of the day. LO 4

Coca-Cola Stock Price Performance

Price per share ($)
46.50

46.00

45.50

45.00

44.50

44.00

43.50

43.00

42.50

42.00

41.50
11-May 12-May 13-May 14-May 15-May 16-May 17-May 18-May 19-May 20-May 21-May 22-May

For which days would you have a gain, and which would you have taken a loss (ignor-
ing transaction costs) for the Coca-Cola data shown in the chart?

Applications

17. Smart Speaker Usage. Futuresource Consulting conducted a survey of owners of

smart speakers to better understand how they use these devices (source: The Wall Street
Journal). The file SmartSpeaker contains the percentage of the respondents who use
SmartSpeaker
their smart speaker for each of 11 activities. LO 1, 2, 4
a. Construct a bar chart that shows the percentage of respondents by category. Use
“How People Use Their Smart Speakers” for the chart title. Edit the chart to make it
easier to interpret. Add data labels.
b. Sort the data so percentages are ordered from smallest to largest, and note the differ-
ences in the corresponding chart.
c. Of the categories in the survey, which is the most popular use of a smart speaker?
What is the least popular use?
18. Age and Ridesharing. A Gallup Poll showed that 30% of all Americans regularly use
a ride-sharing service such as Lyft or Uber. The file RideShare contains the survey
results by age category. LO 1, 2, 4
RideShare
a. Construct a column chart that shows the percentage of respondents who use a ride
share by age category. Use “Who Uses Rideshare?” as the chart title and “Age (years)”
as the horizontal axis title. Edit the chart to make it easier to interpret. Add data labels.
b. Comment on the results.
 Problems 73

19. Exploring Private Colleges (Revisited). In this problem, we revisit the charts in Prob-
lem 11 showing the relationships between tuition and year founded, and graduation rate
and year founded. The two charts are similar. Consider the data in the file Colleges. The
Colleges
file contains the following data for the sample of 102 private colleges: year founded,
tuition and fees (not including room and board), and the percentage of undergraduates
who obtained their degree within six years (source: The World Almanac). LO 1, 2, 4
a. Create a scatter chart to explore the relationship between tuition and percent who
graduate. Use “Graduation Rate versus Tuition” as the chart title, “Tuition” as the
horizontal axis title, and “Graduation Rate (%)” as the vertical axis title.
b. Comment on any apparent relationship.
20. Top Management. The Drucker Institute ranks corporations for managerial effec-
tiveness based on a composite score derived from the following five factors: customer
satisfaction, employee engagement and development, innovation, financial strength,
ManagementTop25
and social responsibility. The file ManagementTop25 contains the top 25 companies in
the Institute’s ranking based on the composite score (source: The Wall Street Journal).
For each company, the industry sector, the company name, and the composite score are
given. LO 1, 2, 4
a. Create a treemap chart using these data with the sector being the category, company
being the subcategory, and the composite score being the quantitative variable.
Use “Management Top 25” for the chart title. Hint: Be sure to first sort the data by
sector.
b. In the sector with the most companies in this top 25, which company has the highest
composite score?
21. Biodiversity Preservation. Ecologists often measure the biodiversity of a region by
the number of distinct species that exist in the region. Nature reserves are lands specif-
ically designated by the government to help maintain biodiversity. Care must be taken
Species
when setting up a network of nature reserves so that the maximum number of species
can exist in the network. Geography matters as well, as putting reserves too close
together might subject the entire network to risks, such as devastation from wildfires.
The initial step in this type of planning usually involves mapping the number of species
that exist in each region. The file Species contains the number of unique species that
exist in each of the 50 states in the United States. LO 1, 2, 4
a. Create a choropleth map that displays number of species by state. Use “Number of
Species per State” for the chart title. Add data labels.
b. Comment on the distribution of species over the United States. Which regions of the
United States have relatively many species? Which regions have relatively few species?
c. Which two states have the most species?
22. Disney Ticket Prices (Revisited). In this problem, we revisit Problem 10, which
displays the price of a general admission ticket to Walt Disney World for the years
2000 to 2020. However, these prices did not factor in inflation over these years. The
DisneyPricesAdjusted
file DisneyPricesAdjusted gives the general admission price and the general admission
price adjusted for inflation for the years 2000 to 2020. LO 1, 2, 4
a. Create a line chart that shows the price of admission and the adjusted price of
admission for the years 2000 to 2020.
b. Explain what the adjusted ticket price data series shows that the nominal ticket price
data series did not.
23. Bubble Chart Labels. The following bubble chart shows TSA wait time (in minutes)
on the horizontal axis, cheapest daily parking rate on the vertical axis, and the size of
each bubble is the number of enplanements in a year (measured in millions). The file
AirportBubbleChart
AirportBubbleChart contains this chart. LO 2, 4
a. Using the following steps, add labels to the bubbles so that the airport codes are eas-
ily identifiable with each bubble.
74 Chapter 2 Selecting a Chart Type

Step 1. Click anywhere on the chart

Step 2. Click the Chart Elements button and select Data Labels
This puts the value of enplanements in each bubble
Step 3. Click the Chart Elements button and move the cursor over Data Labels,
and then select the black triangle on the right Data Labels

Cheapest Parking Rate

$25.00

$20.00

$15.00

$10.00

$5.00

$0.00
0.00 2.00 4.00 6.00 8.00 10.00 12.00
TSA Time

to open a drop-down menu.

From this drop-down menu, select Other Options to reveal the Format
Data Labels task pane
Step 4. When the Format Data Labels task pane appears, under Label Contains,
select Value from Cells and click the Select Range…button
When the Data Label Range dialog box opens, select cells A2:A16 in the
worksheet
Click OK
Step 5. In the Format Data Labels task pane, deselect Y Value under Label Contains
a. in this problem.
b. Which airport has the lowest TSA wait time?
c. Which airport has the most enplanements?
24. Regional Gains and Losses in Population. The United States Univ Census Bureau
tracks shifts in population by each state and region in the United States. The net migra-
tion rate is an indicator of the movement of people from one area to another. A positive
NetMigration
net migration rate means more people moved to the area than moved away, and a
negative net migration rate means more people left the area than moved to the area.
Mathematically the net migration rate N is defined as follows:

(I ] E)
N5 3 1000
M
Where
I = number of people moving to the area in the year under consideration,
E = the number of people moving away from the area in the year under consideration, and
M = the mid-year population of the area. The file NetMigration contains net migration
rates for four regions of the United States. LO 1, 4
 Problems 75

a. Create a heat map using conditional formatting with the Blue-White-Red Color
Scale.
b. Which regions are losing population? Which regions are gaining population?
25. Income Statement. An income statement is a summary of a company’s revenues and
costs over a given period time. The data in the file BellevueBakery is an example of an
income statement. It contains the revenues and costs for last year for Bellevue Bakery.
BellevueBakery
Revenues include gross sales and other income. Costs include returns, cost of goods
sold, advertising, salaries/wages, other operating expenses, and taxes. In the income
statement there are intermediate calculations:
Net Sales = Gross Sales – Returns
Gross Profit = Net Sales – Other Income – Cost of Goods Sold
Net Income Before Taxes = Gross Profit – Advertising – Salaries/Wages – Other
Operating Expenses
Net Income = Net Income Before taxes – Taxes
Create a waterfall chart of the income statement for Bellevue Bakery. Use “Bellevue
Bakery Income Statement” for the chart title. Click the column associated with each
of the calculations above and select Set as Total. Edit the chart to make it easier to
interpret. LO 1, 2
26. Marathon Records. The file MarathonRecords contains marathon world records for
ages from 6 to 90 for women and men (records for ages 9 and 10 are missing). LO 1, 2, 4
a. Create a scatter chart with age on the horizontal axis and the women’s marathon
MarathonRecords
record on the vertical axis. Use “Female Marathon Records (in minutes)” as the ver-
tical axis title and “Age (years)” as the horizontal axis title. Edit the chart to improve
interpretation.
b. Create a scatter chart with age on the horizontal axis and the men’s marathon
record on the vertical axis. Use “Male Marathon Records (minutes)” as the vertical
axis title and “Age (years)” as the horizontal axis title. Edit the chart to improve
interpretation.
c. Create a scatter chart that plots both the women’s record versus age and the men’s
record versus age. Select Scatter with Straight Lines. Use “Marathon Records
(minutes)” as the vertical axis title and “Age (years)” as the horizontal axis title.
Edit the chart to improve interpretation.
d. Based on the charts in parts a, b, and c, what observations can you make regarding
the women’s and men’s marathon records?
Chapter 3
Data Visualization and Design
CONTENTS

DATA VISUALIZATION MAKEOVER 3-3 Data-Ink Ratio

Fast-Food Restaurant Sales 3-4 Other Data Visualization Design Issues
3-1 Preattentive Attributes Minimizing Eye Travel
Color Choosing a Font for Text
Form 3-5 Common Mistakes in Data Visualization
Length and Width Design
Spatial Positioning Wrong Type of Visualization
Movement Trying to Display Too Much Information
3-2 Gestalt Principles Too Many Attributes
Similarity Unnecessary Use of 3D
Proximity Summary
Enclosure
Connection Glossary
Problems

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Define the meaning of preattentive attributes and LO 4 Create data visualizations that are easier for the
explain how preattentive attributes associated with audience to interpret by minimizing the required
color, form, spatial positioning, and movement are eye travel and applying the concepts of preatten-
used in data visualizations tive attributes, Gestalt principles, and decluttering
LO 2 Explain how the Gestalt principles of similarity, LO 5 Explain why certain types of font are preferred over
proximity, enclosure, and connection can be used others for use in text in data visualizations
to create effective data visualizations
LO 6 List several common mistakes in designing data
LO 3 Define data-ink ratio and explain how increasing visualizations and explain how these mistakes can
this ratio through decluttering can create data visu- be avoided
alizations that are easier to interpret
 Data Visualization Makeover 77

D ata V is u ali z ation M a k eover

Fast-Food Restaurant Sales

Charts for making comparisons with data are one of of each fast-food restaurant. Giving a reference value
the most common types of data visualizations created can be effective because it can give the audience a
by analysts. We create charts to compare corporate context to frame the scale of values. However, using
revenues, country populations, student test scores, the shape of the country of Afghanistan here creates
rainfall amounts in different locations, etc. There are several problems. First, it again disguises what is
many ways to show comparisons within a chart, but we being used for comparison since only the height of
must design the chart appropriately so that we do not the shape, and not the overall size of the shape, cor-
confuse the audience or make the chart unnecessarily responds to the GDP. Second, the shape is partially
difficult to interpret. hidden by the logos for McDonald’s and Burger King,
Consider the chart shown in Figure 3.1. This chart making it difficult to see. Finally, it is debatable how
compares the sales of several leading fast-food restau- familiar the audience will be with the actual shape of
rants using the logo of each company to form some- the country of Afghanistan, so it is unclear that includ-
thing similar to a column chart. Using graphics and logos ing this shape provides any additional information or a
can make charts more visually appealing, but we should helpful reference for context to the audience.
make sure that their use does not distract the audi- We also note that this chart shows the vertical axis
ence or make it difficult to correctly interpret the chart. on the right side of the chart, which is unusual. Audi-
Because each logo in Figure 3.1 is two-dimensional (2D) ences typically expect to see the vertical axis on the
(each logo has a length and a width), it is natural to com- left side of the chart. The vertical axis title is also not
pare the sales values based on the overall size, or area, located next to the axis. This requires the audience to
of each logo. However, this is not what is depicted in the move their eyes from place to place on the chart, which
chart. The height of each logo is what is actually being makes it more difficult for the audience to interpret.
used for comparison. The width of each logo is not used Figure 3.2 displays the same data as Figure 3.1,
to convey any meaningful information. To improve the but we have made several changes to the design of
clarity of this chart, it is better to remove the meaning- the chart to make it easier for the audience to inter-
less dimension of the width of the logo in the chart. pret. We have changed the chart to a more typical
There are several other design elements that make column chart that uses the length of the columns
this chart difficult to interpret. The gross domestic rather than the logos to make relative comparisons
product (GDP) for the country of Afghanistan is shown among sales at the fast-food restaurants. We have
here in an attempt to give a relative scale to the sales moved the vertical axis to the left side of the chart,

Figure 3.1 Column Chart Using Logos to Compare Fast-Food Restaurant Sales

$41 BILLION IN SALES

40

35
BILLIONS OF DOLLARS
30
GDP OF AFGHANISTAN
$21 BILLION 25

$4.1 BILLION IN SALES

$11.3 BILLION IN SALES 20
$4.3 BILLION IN SALES $9.4 BILLION IN SALES
$8 BILLION IN SALES 15
$8.2 BILLION IN SALES
10

Source: http://www.princeton.edu/~ina/infographics/starbucks.html
(Continued)
78 Chapter 3 Data Visualization and Design

and we have repositioned the vertical axis label and this makes it clear that this horizontal line is only
above the vertical axis. This makes the chart easier intended to provide reference for the scale of the
for the audience to interpret. We have also used other values. If including the fast-food restaurant
a horizontal line on the column graph to indicate the logos makes the chart more visually appealing to
GDP of Afghanistan. We have made this final change the audience, then these could be included as the
because this value represents a different variable horizontal-axis labels for each column rather than
(GDP) than what is shown on the vertical axis (sales), using the actual names.

Figure 3.2 Improved Column Chart to Compare Fast-Food Restaurant Sales

McDonald’s Sales are Greater than GDP of Afghanistan
Sales ($ billions)
50

40

30
GDP of Afghanistan = $21.0 Billion for comparison
20

10

0
McDonald’s Burger King Wendy’s KFC Pizza Hut Taco Bell Starbucks
Fast Food Restaurants

In this chapter, we discuss specific design elements that can help create effective
data visualizations. However, a good data visualization is not simply created by
following a series of steps. What makes a chart or table effective depends on how well
it addresses the needs of the audience. Therefore, the first step in creating effective
visualizations is understanding the purpose of the chart or table and the needs of the
audience. The design elements covered in this chapter can then be used to most effec-
tively achieve the designated purpose of the visualization and meet the needs of the
audience.
We begin this chapter by discussing preattentive attributes, which are the visual prop-
erties that our minds process without conscious effort. We introduce Gestalt principles,
which explain how people perceive the world around them. We then demonstrate how
preattentive attributes and Gestalt principles can be used to improve data visualizations
through decluttering and increasing the data-ink ratio. We also discuss the importance
of minimizing eye travel required by the audience and using appropriate fonts for text in
a data visualization to make interpretation of visualizations as easy as possible for the
audience. Finally, we conclude the chapter with a discussion of common mistakes made
in data visualization design and how these mistakes can be avoided.

3-1 Preattentive Attributes

The act of “seeing” a chart or table used for data visualization involves a combination of
our eyes and brains. Our eyes receive inputs as reflections of light from a visualization that
our brains then must differentiate and process. The process through which our brains inter-
pret the reflections of light that enter our eyes is known as visual perception.
The process of visual perception is related to how memory works in our brain. At a very
high level, there are three forms of memory that affect visual perception: iconic memory,
 3-1 Preattentive Attributes 79

short-term memory, and long-term memory. Iconic memory is the most quickly processed
form of memory. Information stored in iconic memory is processed automatically, and the
information is held there for less than a second. Short-term memory holds information for
about a minute, and our minds accomplish this by chunking, or grouping, similar pieces of
information together. Estimates vary somewhat, but it is believed that most people can hold
about four chunks of visual information in their short-term memories. For instance, most
people find it difficult to remember which color represents which category if more than
four different colors/categories are used in a bar or column chart. Long-term memory is
We discuss storytelling as it
where we store information for an extended amount of time. Most long-term memories are
relates to data visualization in formed through repetition and rehearsal, but they can also be formed through clever use of
Chapter 7. storytelling.
For most data visualizations, iconic and short-term memory are most important for
visual processing. In particular, an understanding of what aspects of a visualization can
be processed in iconic memory can be helpful for designing effective visualizations.
Preattentive attributes are those features that can be processed by iconic memory. We can
use a simple example to illustrate the power of preattentive attributes in data visualization.
Examine Figure 3.3 and count the number of 7s in the figure as quickly as you can.

FIGURE 3.3 Count the Number of 7s That Appear in This Figure

7 3 4 1 3 4 5 6 4 0
3 0 6 9 0 4 5 8 6 3
2 7 2 2 9 9 4 5 2 1
2 2 4 5 2 0 9 2 0 4
2 4 0 7 6 9 3 0 0 4
7 7 8 9 2 6 7 2 4 7
6 1 3 3 2 1 4 4 9 0
3 6 6 2 7 5 5 2 5 4
1 1 4 0 6 3 4 0 5 1
3 7 5 2 7 5 7 7 3 9
3 3 8 6 9 5 5 3 6 4
7 6 0 3 0 9 9 0 2 9
4 6 9 4 8 2 6 5 8 3
9 3 9 2 2 8 4 3 9 8
5 8 8 2 9 1 2 4 8 5
1 7 4 0 1 1 9 9 5 8

The correct answer is that there are fourteen 7s in this figure. Even if you were able to
find all fourteen 7s, it probably took you quite a bit of time, and it is likely that you made
at least one mistake. Figures 3.4 and 3.5 demonstrate the use of preattentive attributes to
make quickly counting the number of 7s much easier.
Figure 3.4 differentiates the 7s using color; each 7 is orange in color and all other num-
bers are black. This makes it much easier to identify the 7s, and it is much more likely
that you can easily find all fourteen 7s quickly. Figure 3.5 uses size to differentiate the 7s.
Because the 7s are larger than all other numbers, it is much easier to find them quickly.
Both color and size are preattentive attributes; we process these in our iconic memory and
immediately differentiate between their values. This simple example shows the power of
using preattentive attributes to convey meaning in data visualizations.
Proper use of preattentive attributes in a data visualization reduces the cognitive load,
or the amount of effort necessary to accurately and efficiently process the information
being communicated by a data visualization. This makes it easier for the audience to
80 Chapter 3 Data Visualization and Design

FIGURE 3.4 Count the Number of 7s Aided by the Preattentive Attribute

of Color

7 3 4 1 3 4 5 6 4 0
3 0 6 9 0 4 5 8 6 3
2 7 2 2 9 9 4 5 2 1
2 2 4 5 2 0 9 2 0 4
2 4 0 7 6 9 3 0 0 4
7 7 8 9 2 6 7 2 4 7
6 1 3 3 2 1 4 4 9 0
3 6 6 2 7 5 5 2 5 4
1 1 4 0 6 3 4 0 5 1
3 7 5 2 7 5 7 7 3 9
3 3 8 6 9 5 5 3 6 4
7 6 0 3 0 9 9 0 2 9
4 6 9 4 8 2 6 5 8 3
9 3 9 2 2 8 4 3 9 8
5 8 8 2 9 1 2 4 8 5
1 7 4 0 1 1 9 9 5 8

FIGURE 3.5 Count the Number of 7s Aided by the Preattentive Attribute

of Form (Specifically, Size)

7 3 4 1 3 4 5 6 4 0
3 0 6 9 0 4 5 8 6 3

2 7 2 2 9 9 4 5 2 1
2 2 4 5 2 0 9 2 0 4

2 4 0 7 6 9 3 0 0 4

7 7 8 9 2 6 7 2 4 7
6 1 3 3 2 1 4 4 9 0

3 6 6 2 7 5 5 2 5 4
1 1 4 0 6 3 4 0 5 1

3 7 5 2 7 5 7 7 3 9
3 3 8 6 9 5 5 3 6 4

7 6 0 3 0 9 9 0 2 9
4 6 9 4 8 2 6 5 8 3
9 3 9 2 2 8 4 3 9 8
5 8 8 2 9 1 2 4 8 5

1 7 4 0 1 1 9 9 5 8
 3-1 Preattentive Attributes 81

interpret the visualization with less effort. Preattentive attributes related to visual per-
ception are generally divided into four categories: color, form (which includes size),
spatial positioning, and movement. We will examine each of these preattentive attributes
in detail to see how we can use them to reduce cognitive load and create effective data
visualizations.

Color
In terms of data visualization, color includes the attributes of hue, saturation, and lumi-
nance. Figure 3.6 displays the difference between these aspects of color. Hue refers to
what we typically think of as the basis of different colors, for example, red versus blue
versus orange. In technical terms, the hue is defined by the position the light occupies on
the visible light spectrum. Saturation refers to the intensity or purity of the color, which is
defined as the amount of gray in the color. Luminance refers to the amount of black versus
white within the color.

FIGURE 3.6 Hue, Saturation, and Luminance Are All Aspects of the
Preattentive Attribute of Color

Hue

Saturation

Luminance

Hue, saturation, and luminance can each be used to draw the user’s attention to specific
parts of a data visualization and to differentiate among values in a visualization. Using
differences in hue in a data visualization creates bold, stark contracts while changing the
saturation or luminance creates softer, less stark contrasts.
Because color is used
Color can be an extremely effective attribute to use to differentiate particular aspects of
so extensively in data
visualization, we discuss this
data in a visualization. However, one must be careful not to overuse color as it can become
preattentive attribute in much distracting in a visualization. It should also be noted that many people suffer from color­-
more detail in Chapter 4. blindness, which affects their ability to differentiate between some colors.

Form
Form includes the preattentive attributes of orientation, size, shape, length, and width.
Each of these attributes can be used to call attention to a particular aspect of a data visual-
ization. Figure 3.7 shows an example for each of these form-related preattentive attributes.
82 Chapter 3 Data Visualization and Design

FIGURE 3.7 Preattentive Attributes Related to Form

Orientation Size Shape

Length Width

Orientation refers to the relative positioning of an object within a data visualization.

It is a common preattentive attribute present in line graphs. Consider the chart in
Figure 3.8 that visualizes sales of a specific form of syringe that is used for administer-
ing insulin to diabetic patients in Europe and the United States. The difference in the
orientation of these lines makes it easy for the audience to perceive that sales in Europe
are increasing at a much faster rate than the United States for the years 2019 and 2020.

FIGURE 3.8 Demonstration of Preattentive Attribute of Orientation

Sales of Insulin Syringes Increasing Faster in Europe than

in United States
Sales ($ millions)

Europe
2.0

1.5
United States

1.0

2017 2018 2019 2020 2021

Year
 3-1 Preattentive Attributes 83

Because the slope of the line for sales in Europe is much steeper than the slope of the
line for sales in the United States, the orientation of these lines is different. Therefore,
we quickly perceive that sales in Europe have increased much faster than in the United
States since 2019.
Size refers to the relative amount of 2D space that an object occupies in a visualization.
One must be careful with the use of size in data visualizations because humans are not
particularly good at judging relative differences in the 2D sizes of objects. Consider Figure
3.9, which shows a pair of squares and a pair of circles. Try to determine how much larger
the bigger square and bigger circle are than their smaller counterparts.

FIGURE 3.9 Demonstration of Preattentive Attribute of Size: How Much

Larger Are the Bigger Square and Bigger Circle Than Their
Smaller Counterparts?

Both the bigger square and bigger circle are nine times larger than their smaller counter-
parts in terms of area. Most people are not good at estimating this relative size difference,
so we must be careful when using the attribute of size to convey information about relative
amounts.
Avoidance of pie charts is The difficulty most people have in estimating relative differences in 2D size is a major
discussed in Chapter 2. reason why the use of pie charts in a data visualization is generally not recommended.
There are often alternatives to a pie chart that do not rely as heavily on the attribute of size
to convey relative differences in amounts.
Shape refers to the type of object used in a data visualization. Contrary to size and
orientation, the preattentive attribute of shape does not usually convey a sense of quan-
titative amount. In a line graph, the orientation of a line (going up, staying flat, or going
down) generally provides a sense of a quantitative change in amount. For size, most people
assume that a larger object conveys a larger quantitative amount. In general, most shapes
do not specifically correspond to certain quantitative amounts. Nevertheless, shape can be
effectively used to draw attention in a visualization or as a way to group common items
and distinguish between items from different groups.
Figure 3.10 uses the attributes of color and shape to show how items are grouped. For
example, suppose these 20 items represented 20 employees of a company. In Figure 3.10a
we use the preattentive attribute of color to divide the items into three different groups, or
categories: orange, blue, and black. For example, color could represent the type of educa-
tional degree the corresponding employee has earned. Orange could represent a business
degree, blue could represent an engineering degree, and black could represent any other
degree. In Figure 3.10b, we use the preattentive attribute of shape to divide the items into
three groups: circle, square, and triangle. For example, shape could represent the highest
educational degree level that the corresponding employee has earned. A circle could rep-
resent a bachelor’s degree, a triangle could represent a master’s degree, and black could
represent a doctorate degree. In either case, the mind can quickly process these visualiza-
tions and divide the items into their distinct groups. Figure 3.10c uses both attributes, color
and shape, to group items into nine groups—each combination of color (degree type) and
84 Chapter 3 Data Visualization and Design

shape (degree level). It requires a much higher cognitive load here to determine which
items are in the same group. This illustrates why we have to be careful not to overuse
combinations of preattentive attributes, or else we lose the ability for our mind to quickly
recognize these.

FIGURE 3.10 Grouping Items by Color and Shape

(a) (b)

(c)

Length and Width

When we refer to the preattentive attributes of length and width for data visualization, we
are generally referring to their use with lines, bars, or columns. Therefore, length refers to
the horizontal, vertical, or diagonal distance of a line or bar/column while width refers
to the thickness of the line or bar/column (see Figure 3.7). Length is useful for illustrat-
ing quantitative values because a longer line corresponds to a larger value. Length is used
extensively in bar and column charts to visualize data. Because it is much easier to com-
pare relative lengths than relative sizes, bar and column charts are often preferred to pie
charts for visualizing data. Consider data on the number of accounts managed by eight
account managers. Figure 3.11 displays these same data as a pie chart (using size of pie
pieces to indicate number of accounts and color to indicate the manager) and a bar chart
(using length of bars to indicate number of accounts and labels on the vertical axis to indi-
cate the manager).
It is much easier to see that the manager with the most accounts managed is Elijah
and that Kate manages the second-most accounts from the bar chart in Figure 3.11b than
from the pie chart in Figure 3.11a. We could make the bar chart even easier to interpret
 3-1 Preattentive Attributes 85

FIGURE 3.11 Using a Pie Chart (Size and Color) and a Bar Chart (Length) to Display the Accounts
Managed Data

Comparison of Account Mangers Comparison of Account Managers

(by Number of Accounts) Managers
Managers
Sam Kyerstin
Ben Elijah
Megan Kate
Ethan Anna
Anna
Kate Ethan
Elijah Megan
Kyerstin Ben
Sam
0 10 20 30 40 50
Number of Accounts
(a) (b)

by sorting the bars by their length with the longest bar on top and the shortest bar on the
bottom by using the Excel Sort function as described in the following steps. The data
appear in Figure 3.12.

FIGURE 3.12 Data in File AccountsManagedChart

AccountsManagedChart

The following steps applied to the file AccountsManagedChart result in the chart shown
in Figure 3.13.
Step 1. Select cells A1:B9
Step 2. Click the Data tab on the Ribbon
Select Sort
Step 3. When the Sort dialog box appears:
Select the check box for My data has headers
In the Sort by box select Number of Accounts and in the Order box
select Smallest to Largest
Click OK
86 Chapter 3 Data Visualization and Design

FIGURE 3.13 Sorted Bar Chart Comparing Number of Accounts Managed

Comparison of Account Managers

Manager

Elijah

Kate

Megan

Kyerstin

Anna

Ben

Sam

Ethan
0 10 20 30 40 50
Number of Accounts

Line width is used much less frequently in data visualizations. One of the more com-
mon uses of line width in data visualizations is the Sankey chart. A Sankey chart typically
depicts the proportional flow of entities where the width of the line represents the relative
flow rate compared to the widths of the other lines. Figure 3.14 shows an example of a
Sankey chart for the anticipated major and actual major at graduation for students in a
liberal arts college. We can see in Figure 3.14, that most students who anticipate majoring
in the Humanities graduate with a major in Humanities, but we also see that some planned
Humanities majors switch to Social Science majors and fewer switch to Natural Sciences/
Engineering and Interdisciplinary studies. It is relatively easy to interpret which graduation

FIGURE 3.14 Sankey Chart Demonstrating Use of Line Width Attribute

Anticipated Graduation
Major Major

Humanities Humanities

Interdisc Interdisc

Nat Sci/Eng Nat Sci/Eng

Soc Sci

Soc Sci

Undecided

Source: https://www.swarthmore.edu/institutional-research/majors
3-1 Preattentive Attributes 87

major is most popular for a particular anticipated major in Figure 3.14. However, because
it is not easy to compare relative line widths, it is more difficult to compare the proportion
of graduation majors from different anticipated majors. Sankey charts can often quickly
become overwhelming and difficult to interpret, so one should be careful not to try to
include too much information within this type of visualization.

Spatial Positioning
Spatial positioning refers to the location of an object within some defined space.
The spatial positioning most often used in data visualization is 2D positioning. Scatter
charts are a common type of chart that make use of the preattentive attribute of spatial
positioning. Figure 3.15 shows a scatter chart that provides information about the rela-
tionship between annual income and age for a sample of 10 people. Based on the spatial
location of the dots in Figure 3.15, we can easily see based on this sample, that older
people tend to have higher annual incomes and younger people tend to have lower annual
incomes.

FIGURE 3.15 Scatter Chart Demonstrating Spatial Positioning Attribute

Comparing Income and Age

Annual Income ($)
120,000

100,000

80,000

60,000

40,000

20,000

–
20 25 30 35 40 45 50
Age (years)

Movement
Humans are attuned to detecting movement. Therefore, preattentive attributes such as
flicker and motion can be effective at drawing attention to specific items or portions of a
data visualization. Flicker refers to effects such as flashing to draw attention to something,
while motion involves directed movement and can be used to show changes within
a visualization. Because many tables and charts are static, movement is not possible in
many contexts of data visualization. However, when the data visualization tool allows for
the use of movement, it can be used to direct attention to certain areas of a visualization
or to show changes over time or space. Because the focus of this text is on static visual-
izations, we will not go into detail on using the preattentive attribute of movement, but we
88 Chapter 3 Data Visualization and Design

should caution that movement can also become overwhelming and distracting if it is over-
used in a visualization.

3-2 Gestalt Principles

Gestalt is a German word Gestalt principles refer to the guiding principles of how people interpret and perceive
meaning “form.” The Gestalt what they see. These principles can be used in the design of effective data visualiza-
principles are based on ideas
tions. The principles generally describe how people define order and meaning in things
that date back to experiments
conducted in the early 20th
that they see. We will limit our discussion to the four Gestalt principles that are most
century by Max Wertheimer, closely related to the design of data visualizations: similarity, proximity, enclosure,
Kurt Koffka, and others to and connection. An understanding of these principles can help in creating more effec-
examine how people perceive tive data visualizations and help differentiate between clutter and meaningful design in
the world around them.
data visualizations.

Similarity
The Gestalt principle of similarity states that people consider objects with similar charac-
teristics as belonging to the same group. These characteristics could be color, shape, size,
orientation, or any preattentive attribute. When a data visualization includes objects with
similar characteristics, it is important to understand that this communicates to the audience
that these objects should be seen as belonging to the same group. Figure 3.16 is a portion
of what was shown in Figure 3.10, but here we are using it to represent the Gestalt princi-
ple of similarity. The audience will perceive objects that are the same color, or same shape,
as belonging to the same group. We need to understand this when we design a visualization
and make sure that we only use similar characteristics for objects when they belong to the
same group.

FIGURE 3.16 Illustrating the Gestalt Principle of Similarity

Proximity
The Gestalt principle of proximity states that people consider objects that are physically
close to one another as belonging to a group. People will generally seek to collect objects
that are near each other into a group and separate objects that are far from one another into
different groups. The principle of proximity is apparent in many data visualization charts,
including scatter charts.
Consider a firm that would like to perform a market segmentation analysis of its cus-
tomers to learn more about the customers who purchase its products. The company has
collected data on the ages and annual incomes of its customers. A simple scatter chart of
the age and income of customers is shown in Figure 3.17. Here, our natural inclination is to
 3-2 Gestalt Principles 89

view this as three distinct groups of customers based on the proximity of the points. This is
an example of the Gestalt principle of proximity.

FIGURE 3.17 Scatter Chart Illustrating Gestalt Principle of Proximity

Marketing Segmentation by Customer Age and Income

Annual Income ($)

140,000

120,000

100,000

80,000

60,000

40,000

20,000

–
0 10 20 30 40 50 60 70
Age (years)

Enclosure
To create enclosures such The Gestalt principle of enclosure states that objects that are physically enclosed
as the ones in Figure 3.18 together are seen as belonging to the same group. We can illustrate this princi-
in Excel, click Insert on the
Ribbon, then click Shapes in
ple using two modified versions of Figure 3.17. First, we can simply reinforce the
the Illustrations group. similarity principle by creating an enclosure of the points that are already in close
proximity (see Figure 3.18a). Alternatively, suppose that there is a third attribute of
the customers, other than annual income and age, which can be used to group these
customers such as educational background. If we want to visually indicate certain
customers that share this characteristic of having similar educational backgrounds,
then we can use the principle of enclosure to illustrate this even when customers do
not appear close together in the chart. This is shown in Figure 3.18b. Note that the
enclosure can be indicated in multiple ways in a chart. In Figure 3.18a we have used
shaded areas to enclose points. In Figure 3.18b we have used dashed boxes. In gen-
eral, we only need to create a suggestion of enclosure for the audience to view the
objects being enclosed as members of the same group.

Connection
The Gestalt principle of connection states that people interpret objects that are con-
nected in some way as belonging to the same group. One of the most common uses
of the principle of connection in data visualization is for time-series data. Consider a
data center company that wants to compare its forecast to actual server loads from cus-
tomers over the past 14 days. Figure 3.19a shows the company’s forecasts and actual
values of peak server loads (in terms of requests per second) for the past 14 days.
Compare Figure 3.19a to Figure 3.19b.
90 Chapter 3 Data Visualization and Design

FIGURE 3.18 Scatter Charts Illustrating Gestalt Principle of Enclosure

Marketing Segmentation by Customer Age and Income

Annual Income ($)

140,000

120,000

100,000

80,000

60,000

40,000

20,000

–
0 10 20 30 40 50 60 70
Age (years)
(a)

Marketing Segmentation by Customer Age and Income

Annual Income ($)

140,000

120,000

100,000

80,000

60,000

40,000

20,000

–
0 10 20 30 40 50 60 70
Age (years)
(b)

Whether or not to retain the Figure 3.19b connects the markers in each series. Connecting these markers makes any
markers for the individual data
trends in the data much more obvious and it becomes easier to separate the forecast values
points in a time-series chart is
mostly a matter of preference.
from the actual values. Because the time-series data are connected, the audience interprets
Here, we have retained the these points as belonging to the same group, and patterns become more obvious. In Figure
markers, but they could also 3.19b, the principle of connection makes it easier for the audience to see that there appears
be removed to show only be some sort of repeating pattern in our data where server loads increase for several days
the line.
3-3 Data-Ink Ratio 91

FIGURE 3.19 Chart for Time-Series Data Illustrating Gestalt Principle of

Connection

Server Load Forecast vs. Actual For Past 14 Days

Actual
Peak Server Load (requests per second)
Forecast
35
30
25
20
15
10
5
0
0 2 4 6 8 10 12 14
Day
(a)

Server Load Forecast vs. Actual For Past 14 Days

Peak Server Load (requests per second) Actual
35 Forecast

30
25
20
15
10
5
0
0 2 4 6 8 10 12 14
Day
(b)

and then decrease, and also that our forecast is consistently over-forecasting peak demand
on the days with lowest demand.

3-3 Data-Ink Ratio

The concepts of preattentive attributes and Gestalt principles are valuable in understand-
ing features that can be used to visualize data and how visualizations are processed by
the mind. However, it is easy to overuse any of the features and diminish the effective-
ness of the feature to differentiate and draw attention. A guiding principle for effective
data visualizations is that the table or graph should illustrate the data to help the audience
generate insights and understanding. The table or graph should not be so cluttered as to
disguise the data or be difficult to interpret.
92 Chapter 3 Data Visualization and Design

A common way of thinking about this principle is the idea of maximizing the data-ink
The data-ink ratio was ratio. The data-ink ratio measures the proportion of “data-ink” to the total amount of
introduced by Edward Tufte ink used in a table or chart, where data-ink is the ink used that is necessary to convey the
in his 1983 book The Visual
meaning of the data to the audience. Non-data-ink is ink used in a table or chart that serves
Display of Quantitative
Information. no useful purpose in conveying the data to the audience. Note in Figure 3.11a that the pie
chart uses color and a legend to differentiate between the eight managers. The bar chart in
this figure communicates the same information without either of these features, and so has
a higher data-ink ratio.
Let us consider the case of Diaphanous Industries, a firm that produces fine silk cloth-
ing products. Diaphanous is interested in tracking the sales of one of its most popular
items, a particular style of scarf. Table 3.1 and Figure 3.20 provide examples of a table
and chart with low data-ink ratios used to display sales of this style of scarf. The data
used in this table and figure represent product sales by day. Both of these examples are
similar to tables and charts generated with Excel using common default settings. In Table
3.1, most of the gridlines serve no useful purpose. Likewise, in Figure 3.20, the grid-
lines in the chart add little additional information. In both cases, most of these lines can
be deleted without reducing the information conveyed. However, an important piece of
information is missing from Figure 3.20: titles for axes. Generally, axes should always be
labeled in a chart. There are rare exceptions to this where both the meaning and unit of
measure are obvious such as when the axis displays the names of months (i.e., “January,”
“February,” “March,” etc.). For most charts, we recommend labeling the axes to avoid
the possibility of misinterpretation by the audience and to reduce the cognitive load
required by the audience.

TABLE 3.1 Example of Low Data-Ink Ratio Table

Scarf Sales
Day Sales (units) Day Sales (units)
1 150 11 170
2 170 12 160
3 140 13 290
4 150 14 200
5 180 15 210
6 180 16 110
7 210 17 90
8 230 18 140
9 140 19 150
10 200 20 230

Table 3.2 shows a modified table in which the gridlines have been deleted. Deleting the
gridlines in Table 3.1 increases the data-ink ratio because a larger proportion of the ink in
the table is used to convey the information (the actual numbers). Similarly, deleting the
unnecessary horizontal and vertical gridlines in Figure 3.20 increases the data-ink ratio.
Note that deleting these gridlines and removing (or reducing the size of) the markers at
each data point can make it more difficult to determine the exact values plotted in the
chart. Thus, understanding the needs of the audience is essential to defining what consti-
tutes a good visualization for that audience. If the audience needs to know the exact sales
values on different days, then it may be appropriate to add ink to the chart to label each
data point, or even to display the data as a table, rather than as a chart.
 3-3 Data-Ink Ratio 93

FIGURE 3.20 Example of Low Data-Ink Ratio Chart

Scarf Sales
Sales (units)
350

300

250

200

150

100

50

0
0 5 10 15 20

In many cases, white space, the portion of a data visualization that is devoid of markings,
can improve readability in a table or chart. This principle is similar to the idea of
increasing the data-ink ratio. Consider Table 3.2 and Figure 3.21. Removing the unnec-
essary lines has increased the white space, making it easier to read both the table and the
chart. The fundamental idea in creating effective tables and charts is to make them as
simple as possible in conveying information to the reader. The following steps describe
how to modify the chart in Figure 3.20 to appear as it does in Figure 3.21.

TABLE 3.2 Increasing the Data-Ink Ratio in a Table by Removing

Unnecessary Gridlines
Scarf Sales
Day Sales (units) Day Sales (units)
1 150 11 170
2 170 12 160
3 140 13 290
4 150 14 200
5 180 15 210
6 180 16 110
7 210 17 90
8 230 18 140
9 140 19 150
10 200 20 230

Step 1. Click anywhere on the chart in the file ScarfSalesChart

Click the Chart Elements button
ScarfSalesChart Deselect the check box for Gridlines
Deselect the check box for Legend
94 Chapter 3 Data Visualization and Design

The Chart Elements button is Step 2. Double-click one of the data points in the chart
not available in Mac versions
When the Format Data Series task pane appears, click the Fill & Line
of Excel. See the NOTES +
COMMENTS at the end of
icon
Section 2-2 for a description of Select Marker and under Marker Options select None
how to access these features Step 3. Click the chart title “Scarf Sales”
on a Mac. Drag this text box to the left so that it aligns with the vertical axis
Click the Home tab on the Ribbon and select the Align Left button in
the Alignment group to left justify the chart title
Click the chart title text box, type Sales (units) under the existing chart
title, and change the font of this text to Calibri 10.5 Bold to create the
vertical axis title (see Figure 3.21)
A
Step 4. Click the Insert tab on the Ribbon and select Text Box from the Text group Text

Click on the chart just below the end of the horizontal axis to create a title
for this axis
Type Day into the text box and position it just below the end of the hori-
zontal axis
Change the font for this text to Calibri 10.5 Bold to create the horizontal
axis title (see Figure 3.21)

FIGURE 3.21 Increasing the Data-Ink Ratio by Adding Axis Title and
Removing Unnecessary Lines and Labels

Scarf Sales
Sales (units)
350

300

250

200

150

100

50

0
0 5 10 15 20
Day

Comparing Figure 3.20 to Figure 3.21 also illustrates the importance of editing the default
charts created in Excel. Without additional edits, most charts created in Excel will appear
more like Figure 3.20 than Figure 3.21. It is important to spend the extra time to add axis
titles where needed, remove unnecessary gridlines, and format the chart title to improve
the data-ink ratio of charts created in Excel. This will greatly enhance the visual appeal of
these charts and make them easier for the audience to interpret.
The process of increasing the data-ink ratio in charts is also known as decluttering.
Decluttering refers to removing the clutter, or non-data-ink, in a chart. In Figure 3.20,
the gridlines are considered clutter as they provide little value to the audience’s ability to
interpret the chart.
 3-3 Data-Ink Ratio 95

As another example of decluttering, consider Figure 3.22. This figure shows the course
evaluations for Professor Bob Smith’s Statistics (STAT) 7011 course, Introduction to
Analytics. The chart shows Professor Smith’s course evaluation score for the question,
“How effective is this instructor?” which is scored on a scale of 1 = Not effective to
5 = Extremely effective. The historical average score over all courses in the college for this
question is 3.75, so this value is also indicated on the chart.

FIGURE 3.22 A Cluttered Column Chart for the STAT 7011 Course Evaluations Data

Professor Smith STAT 7011 Course Evaluations

How effective is this instructor? (1 = Not effective; 5 = Extremely effective) College Average = 3.75
5.0
4.3 4.3
4.5 4.1 4.2 4.2
3.8 3.9 4.1 4.0 3.7 4.0
4.0 3.7 3.6
3.5
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Spring Fall Spring Fall Spring Fall Spring Fall Spring Fall Spring Fall Spring Fall
2014 2014 2015 2015 2016 2016 2017 2017 2018 2018 2019 2019 2020 2020
Year

The chart in Figure 3.22 has a low data-ink ratio and is cluttered. We can improve the data-
ink ratio and declutter the chart by removing aspects of the chart that are not helpful to the
audience. We can remove the gridlines and simplify the horizontal axis. By adding a legend
that specifies Fall Semester and Spring Semester, we can simplify the horizontal axis by
only showing years. We can simplify the vertical axis by removing some of the markings
to increase the use of white space. If our main goal for the chart is to compare Professor
Smith’s performance to the college average, we can also remove the data labels from each
column to further declutter the chart.
The steps below show how to declutter this chart to improve the data-ink ratio. Because
Excel is somewhat limited in the editing functions available for charts to create the most
effective designs, we will manually adjust several elements of this chart using text boxes to
declutter the chart.
Step 1. Click anywhere on the chart in the file CourseEvalsChart
Click the Chart Elements button
Deselect the check box for Gridlines
CourseEvalsChart
Deselect the check box for Data Labels
Step 2. Click anywhere on the chart
Select the vertical axis
Step 3. When the Format Axis task pane appears:
Change the entry in the Major Box under Units from 0.5 to 1.0 to
remove the vertical-axis increments of 0.5
96 Chapter 3 Data Visualization and Design

Step 4. Double-click the second column in the chart corresponding to “Fall 2014” so
that only this column is selected in the chart
In the Format Data Point task pane, click the Fill & Line icon
Under Fill, select a darker blue color for this column
Step 5. Repeat Step 4 for each column corresponding to a Fall semester in the chart
(see Figure 3.23)
Step 6. Click the semester labels below the horizontal axis and press the Delete key to
remove the name of each column
Step 7. Delete the text in cells A2:A15 to remove the horizontal axis labels.
Type 2014 in cell A2, 2015 in cell A4, 2016 in cell A6, 2017 in cell A8,
2018 in cell A10, 2019 in cell A12 and 2020 in cell A14 to add the years
to the horizontal axis on the chart.
Step 8. Click the arrow in the chart pointing to the line for the college average evalua-
tion score and press the Delete key
Step 9. Move the text box containing College Average = 3.75 closer to the line that
marks the college average evaluation score
Step 10. Click the Insert tab on the Ribbon
A
Click Text Box in the Text group and click near the upper right of the
Text
chart Box

Wingdings is a font that Click the Home tab in the Ribbon and change the font type to
contains a variety of symbols Wingdings 10.5
that can be edited as text. The
Type n in the text box to create a square box
Wingdings font can be useful
to create text that matches
Change the font color of the square box to match the lighter blue columns
the symbols used in Excel for in the chart
creating charts. Change the font type back to Calibri 10.5 and the font color back
to black
Type Spring Semester and press Enter
Click the Home tab in the Ribbon and change the font type to
Wingdings 10.5
Type n in the text box to create a square box
Change the font color of the square box to match the darker blue columns
in the chart
Change the font type back to Calibri 10.5 and the font color back to black
Type Fall Semester (see the legend in Figure 3.23)
Step 11. Line up any added elements of the chart that may have been moved in the edit-
ing process. See Figure 3.23 for how the finished chart should appear.
Figure 3.23 has a higher data-ink ratio and more white space than Figure 3.22, which
makes it easier for the audience to interpret. The vertical and horizontal axis labels are now
cleaner, and redundant information has been removed. By differentiating Fall Semester and
Spring Semester using the preattentive attribute of color and manually creating a legend,
we also make it easier for the audience to compare Fall Semester evaluation scores to each
other and Spring Semester evaluation scores to each other.
Decluttering is also applicable to tables used in data visualization. In keeping with the
principle of maximizing the data-ink ratio, for tables this usually means avoiding vertical
lines in a table unless they are necessary for clarity. Horizontal lines are generally only
necessary for separating column titles from data values or when indicating that a calcu-
lation has taken place. Consider Figure 3.24, which compares several forms of a table
displaying cost, revenue, and profit data for a company. Most people find Design D, with
the fewest gridlines, easiest to read. In this table, gridlines are used only to separate the col-
umn headings from the data and to indicate that a calculation has occurred to generate the
Profits row and the Total column.
 3-3 Data-Ink Ratio 97

FIGURE 3.23 Improved Column Chart through Decluttering for the STAT 7011 Course
Evaluations Data

Professor Smith STAT 7011 Course Evaluations

How effective is this instructor? (1 = Not effective; 5 = Extremely effective)
5.0 Spring Semester
Fall Semester

4.0 College Average =3.75

3.0

2.0

1.0

0.0
2014 2015 2016 2017 2018 2019 2020
Ye ar

FIGURE 3.24 Comparing Different Table Designs with Different Data-Ink Ratios

Design A: Design C:
Month Month
1 2 3 4 5 6 Total 1 2 3 4 5 6 Total
Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 326,567 Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 326,567
Revenues ($) 64,124 66,128 67,125 48,178 51,785 55,687 353,027 Revenues ($) 64,124 66,128 67,125 48,178 51,785 55,687 353,027
Profits ($) 16,001 9,670 3,000 (3,980) (2,933) 4,702 26,460 Profits ($) 16,001 9,670 3,000 (3,980) (2,933) 4,702 26,460

Design B: Design D:
Month Month
1 2 3 4 5 6 Total 1 2 3 4 5 6 Total
Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 326,567 Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 326,567
Revenues ($) 64,124 66,128 67,125 48,178 51,785 55,687 353,027 Revenues ($) 64,124 66,128 67,125 48,178 51,785 55,687 353,027
Profits ($) 16,001 9,670 3,000 (3,980) (2,933) 4,702 26,460 Profits ($) 16,001 9,670 3,000 (3,980) (2,933) 4,702 26,460

In large tables, we can use vertical lines or light shading to help the audience differen-
tiate the columns or rows. Table 3.3 displays revenue data by location for nine cities
and shows 12 months of revenue and cost data. In Table 3.3, every other column has
been lightly shaded. This helps the reader quickly scan the table to see which values
correspond with each month. The horizontal line between the revenue for Academy and
the Total row helps the audience differentiate the revenue data for each location and
indicates that a calculation to generate the totals by month has been performed. If one
wanted to highlight the differences between locations, we could shade every other row
instead of every other column.
98 Chapter 3 Data Visualization and Design

Notice also the alignment of the text and numbers in Table 3.3. Columns of numerical
values in a table should usually be right-aligned; that is, the final digit of each number
should be aligned in the column. This makes it easy to see differences in the magnitude
of values. If you are showing digits to the right of the decimal point, all values should
include the same number of digits to the right of the decimal. Also, use only the number
of digits that are necessary to convey the meaning in comparing the values; including
additional digits that are not meaningful for comparisons to the audience increases clut-
ter. In many business applications, we report financial values, in which case we often
round to the nearest dollar or include two digits to the right of the decimal when such
precision is necessary. For extremely large numbers, we may prefer to display data
rounded to the nearest thousand, ten thousand, or even million. For instance, if we need
to include, say, $3,457,982 and $10,124,390 in a table when exact dollar values are not
necessary, we could write these as 3.458 and 10.124 and indicate that all values in the
table are in units of $1,000,000 (or $ millions).
It is generally most effective to left-align text values within a column in a table, as in
the Revenues by Location (the first) column of Table 3.3. In some cases, you may prefer
to center text, but you should do this only if the text values are all approximately the same
length. Otherwise, aligning the first letter of each data entry promotes readability. Column
headings should either match the alignment of the data in the columns or be centered over
the values, as in Table 3.3.

Table 3.3 Table Showing Revenues by Location for 12 Months of Data

Month
Revenues by
Location ($) 1 2 3 4 5 6 7 8 9 10 11 12 Total
Temple 8,987 8,595 8,958 6,718 8,066 8,574 8,701 9,490 9,610 9,262 9,875 11,058 107,895
Killeen 8,212 9,143 8,714 6,869 8,150 8,891 8,766 9,193 9,603 10,374 10,456 10,982 109,353
Waco 11,603 12,063 11,173 9,622 8,912 9,553 11,943 12,947 12,925 14,050 14,300 13,877 142,967
Belton 7,671 7,617 7,896 6,899 7,877 6,621 7,765 7,720 7,824 7,938 7,943 7,047 90,819
Granger 7,642 7,744 7,836 5,833 6,002 6,728 7,848 7,717 7,646 7,620 7,728 8,013 88,357
Harker Heights 5,257 5,326 4,998 4,304 4,106 4,980 5,084 5,061 5,186 5,179 4,955 5,326 59,763
Gatesville 5,316 5,245 5,056 3,317 3,852 4,026 5,135 5,132 5,052 5,271 5,304 5,154 57,859
Lampasas 5,266 5,129 5,022 3,022 3,088 4,289 5,110 5,073 4,978 5,343 4,984 5,315 56,620
Academy 4,170 5,266 7,472 1,594 1,732 2,025 8,772 1,956 3,304 3,090 3,579 2,487 45,446
Total 64,124 66,128 67,125 48,178 51,785 55,687 69,125 64,288 66,128 68,128 69,125 69,258 759,079
Costs ($) 48,123 56,458 64,125 52,158 54,718 50,985 57,898 62,050 65,215 61,819 67,828 69,558 710,935

3-4 Other Data Visualization Design Issues

Minimizing Eye Travel
Data visualizations should be easy to view and interpret by the audience. Charts and tables
should reveal insights to the audience, while minimizing the cognitive load required of
the audience. We can minimize cognitive load by using preattentive attributes and Gestalt
principles as well as by increasing the data-ink ratio in our data visualizations. We can also
minimize cognitive load by minimizing the eye travel required by the audience.
Consider the Office of Budget and Performance Improvement for the City of
Springfield. This city office would like to compare the performance of the two police dis-
tricts located in its city. One performance metric used by the city is clearance rate, which is
the fraction of reported crimes that result in an arrest. Figure 3.25 compares the clearance
rates for property crimes in Springfield’s District 1 and District 2 over the last 6 months.
 3-4 Other Data Visualization Design Issues 99

FIGURE 3.25 A Default Line Chart Created in Excel for Property Crime
Clearance Rates

Property Crime Clearance Rates in Springfield

30

Proportion Closed with Arrest (%)

25

20

15

10

0
1 2 3 4 5 6
Month
District 1 District 2

Figure 3.25 has several characteristics that increase the required eye travel for the audience.
Many of these characteristics are typical of default charts created in Excel. First, the leg-
end is located at the bottom of the chart. This requires the audience members to look at the
legend at the bottom of the chart and then move their eyes up to the lines to match the line
type from the legend with the correct line in the chart. We can greatly reduce the eye travel
required of the audience by moving the legend closer to the lines or, even better, by directly
labeling each line in the chart. Second, Excel also typically inserts vertical-axis title text as
rotated 90 degrees from the chart title and horizontal-axis title. This requires the audience’s
eyes to move all around the chart to read the horizontal-axis title, the vertical-axis title, and the
chart title. It is better to align the titles within a chart as much as possible so the audience can
look at only a few places to quickly interpret the chart. The steps below demonstrate how we
can improve Figure 3.25 to reduce the amount of eye travel required by the audience.
Step 1. Click anywhere on the chart in the file ClearanceRatesChart
Click the Chart Elements button
Deselect the check box for Legend
ClearanceRatesChart
Step 2. Double-click the last data point on the line in the chart for the District 1 data
to select only that data point
Right-click the selected data point and select Add Data Label (this will
add a data label with the value of this data point which is “24”)
Change the “24” in this data label to District 1 and change the font to
Calibri 10.5
Step 3. Double-click the last data point on the line in the chart for the District 2 data
to select only that data point
Steps 4–6 add labels to the
Right-click the selected data point and select Add Data Label (this will
lines within the chart so that
a separate legend is not
add a data label with the value of this data point, which is “20”)
necessary to minimize eye Change the “20” in this data label to District 2 and change the font to
travel. Calibri 10.5
Step 4. Click the vertical-axis title “Proportion Closed with Arrest (%)”
Press the Delete key
100 Chapter 3 Data Visualization and Design

Excel’s automatic axis title Step 5. Click the chart title “Property Crime Clearance Rates in Springfield”
options only allow for Drag this text box to the left so that it aligns with the vertical axis
limited formatting changes.
Click the Home tab on the Ribbon and select the Align Left button in
Therefore, to create a vertical-
axis title above the axis that
the Alignment group to left justify the chart title
minimizes eye travel, we Step 6. Click the Insert tab on the Ribbon
A
manually create this with a
Click Text Box Text in the Text group and click above the vertical axis to
text box. Box

create a text box for the vertical axis title

Type Proportion Closed with Arrest (%) in the text box and change the
font to Calibri 10.5 Bold to create the vertical axis title (see Figure 3.26)
Step 7. Click the horizontal-axis title “Month” and drag the text box to the end of the
horizontal axis so that it lines up with the last label on the horizontal axis, in
this case the “6” (see Figure 3.26)

FIGURE 3.26 Improved Line Chart for Property Crime Clearance Rates to
Reduce Required Eye Travel of Audience

Property Crime Clearance Rates in Springfield

Proportion Closed with Arrest (%)
30

25
District 1

20 District 2

15

10

0
1 2 3 4 5 6
Month

Figure 3.26 reduces the eye travel required of the audience by aligning the chart title
and vertical-axis title with the vertical axis, moving the line labels adjacent to the chart
lines, and aligning the horizontal-axis label with the line labels and the end of the horizon-
tal axis. This allows the audience to naturally view the chart by scanning from left to right
and easily interpreting all important information with low cognitive load. This is the final
form that we will use for most charts in this textbook. The preceding steps require addi-
tional effort from what is easily created in Excel, but we recommend these steps when cre-
ating a final version of a data visualization to make it as easy as possible for the audience
to interpret your chart.

Choosing a Font for Text

Text is an important part of any data visualization. It is used to label axes, fill in table val-
ues, and call out important aspects of the visualization to the audience. Because text is such
an important part of a data visualization, the font that is used to display the text is also an
important consideration. Most data visualization software tools, including Excel, allow the
user to choose from dozens, and even hundreds, of font options for displaying text.
 3-4 Other Data Visualization Design Issues 101

Not all data visualization experts agree on the preferred type of font to use for text in data
visualizations, and often this choice may depend on the needs of the audience or on other design
elements of the data visualization. However, most experts agree that some font types are gen-
erally preferred for text in a data visualization over others; for example, sans-serif fonts (fonts
that do not contain serifs) are generally preferred over serif fonts (fonts that do contain serifs)
for text in a data visualization. Serifs refer to the small end-of-stroke features that are visual
in the characters created using serif fonts. Figure 3.27 illustrates the difference between sans-
serif and serif fonts. Common serif fonts include Times, Times New Roman, and Courier.
Common sans-serif fonts include Arial, Calibri, Myriad Pro and Verdana.
In general, serif fonts are preferred for printed work and sans-serif fonts are preferred for
text displayed digitally. Sans-serif fonts are also often more legible than serif fonts at small
sizes. Because data visualizations are often viewed in both print form and digitally, and
because data visualizations often contain fonts of many different sizes, sans-serif fonts are
generally preferred over serif fonts for text in data visualizations. In this textbook, all charts
provided in Excel use the sans-serif font Calibri because it is the default font in Excel. Most
printed charts in the textbook use the sans-serif font Myriad Pro because it is legible at many
different sizes and it works well for both print and digital work. However, other sans-serif
fonts, such as Arial and Verdana, are also usually acceptable for data visualization purposes.

FIGURE 3.27 Serif versus Sans-Serif Font for Text

Serif Font Sans-Serif Font

F Serifs R F R

Perhaps even more important than whether text in a data visualization uses serif or sans-
serif font is that the choice of fonts is consistent and restrained. Generally, charts should
use the same font type and same font size when conveying similar information to the audi-
ence. For instance, horizontal- and vertical-axes titles should generally have the same type
and size of font. Because size is a preattentive attribute, the use of inconsistent font sizes
for similar information will direct audience attention to certain parts of a visualization and
will immediately be processed as having additional meaning. The use of too many types of
fonts can create clutter and increase the cognitive load of interpreting the visualization for
the audience. Therefore, most data visualizations should use a single type of font for text,
and make use of different font sizes as well as bold, italic, and possibly color enhance-
ments to differentiate between features and direct audience attention.

N otes 1 C omments

1. The chart design steps we illustrate in this section require 2. When to include axes titles and chart titles is often a mat-
extra work. Whether this extra work is necessary depends ter of opinion. We recommend having a chart title when
on the intended use and audience for the chart. If the chart it is necessary for clarification or to make explicit the
is for simply exploring data or for use as a draft to generate intended message of the chart. We recommend omitting
feedback, it is less important to format all elements; refor- chart and axes titles only when they provide redundant
mats such as changing axes label locations and changing information.
font sizes of text are not necessary at that point. However, for 3. Other software packages commonly used for data visualiza-
final versions of charts that will be presented to an external tion, such as Tableau, Power BI, and R, have more flexibility
audience, it is often worthwhile to spend extra time to make for formatting charts than Excel, but they also require more
the chart as easy as possible for the audience to interpret. time to learn their full functionality.
102 Chapter 3 Data Visualization and Design

3-5 Common Mistakes in Data Visualization Design

Wrong Type of Visualization
The best type of chart or table to use for data visualization strongly depends on the audi-
ence that will view the visualization as well as the insights or story that is to be told
through the visualization. Throughout this textbook, we provide best practices for design-
ing effective data visualizations, but many of the decisions related to which chart to use and
some aspects of the design will depend on the situation and goal of the visualization. In
this section, we use the concepts presented in this chapter to discuss several situations for
Understanding the needs of
which one type of visualization is preferred over another. However, we must keep in mind
the audience is discussed in that the most effective visualization depends on the needs of the audience and the message
more detail in Chapter 7. we are trying to convey.
If the goal of the visualization is to convey precise numerical values, then it is often
preferable to use a table rather than a chart. Because it is more difficult for an audience to
Recommendations for when to
make relative comparisons on the preattentive attribute of shape than on the preattentive
use a table versus a chart and
for which type of chart to use
attribute of length, bar or column charts are generally preferred over pie charts. However,
are discussed in more detail in there are cases for which the most appropriate type of visualization depends on the goal of
Chapter 2. the visualization and is not always obvious.
Consider the case of Stanley Consulting Group, a company that provides analytics
consulting to nonprofit companies. Stanley Consulting Group has offices in Hartford,
Stamford, and Providence. Each office has a similar number of consultants and similar
performance expectations. Stanley Consulting Group would like to compare the per-
formance of each office. It is mostly interested in comparing each office’s performance
relative to the quarterly goal, and in identifying trends over time at each location.
Figure 3.28 uses a clustered column chart to compare the performances of the offices
in terms of quarterly booked revenue for the previous six quarters. The chart also com-
pares this performance to the quarterly booked revenue goal of $600,000 that applies to
each office.
This clustered column chart in Figure 3.28 is not necessarily inappropriate for any obvi-
ous reason, but we should try to visualize the same data with other charts to see if we can

FIGURE 3.28 A Clustered Column Chart for Booked Revenue Data

Comparison at Stanley Consulting Group

Booked Revenue Comparison

($1000s)
Hartford Stamford Providence
800

700
Goal = $600,000
600

500

400

300

200

100

0
1 2 3 4 5 6
Quarter
 3-5 Common Mistakes in Data Visualization Design 103

improve on this design. The steps below show how we can change the clustered column
chart shown in Figure 3.28 to a line chart in Excel.
Step 1. Right-click anywhere on the chart in the file BookedRevenueChart
Select Change Chart Type...
Step 2. When the Change Chart Type dialog box appears
BookedRevenueChart
Select Line
Click OK
Step 3. Click anywhere on the line chart
Click the Chart Elements button
Deselect the check box for Legend
Step 4. Double-click the last data point on the line in the chart for the Stamford data
Steps 4–6 add labels to the to select only this data point
lines within the chart so that Right-click the selected data point and select Add Data Label (this will
a separate legend is not
add a data label with the value of this data point, which is “665”)
necessary to minimize eye
travel. Change the “665” in this data label to Stamford and change the font to
Calibri 10.5
Step 5. Double-click the last data point on the line in the chart for the Providence data
to select only that data point
Right-click the selected data point and select Add Data Label (this will
add a data label with the value of this data point, which is “575”)
Change the “575” in this data label to Providence and change the font to
Calibri 10.5
Step 6. Double-click the last data point on the line in the chart for the Hartford data to
select only that data point
Right-click the selected data point and select Add Data Label (this will
add a data label with the value of this data point, which is “420”)
Change the “420” in this data label to Hartford and change the font to
Calibri 10.5
These steps result in the line chart shown in Figure 3.29.

FIGURE 3.29 A Line Chart for the Booked Revenue Data Comparison

Booked Revenue Comparison

($1000s)
800

700
Goal = $600,000 Stamford
600
Providence
500

Hartford
400

300

200

100

0
1 2 3 4 5 6
Quarter
104 Chapter 3 Data Visualization and Design

Using a line chart for these data in Figure 3.29 has several advantages. First, the Gestalt
principle of connectedness in the line chart makes it much easier to see trends in booked
revenue by office. From Figure 3.29 we see that Hartford exceeded the booked reve-
nue goal in Quarter 1, but that it has fallen below the goal consistently after that and its
booked revenues are generally declining. Booked revenues at the Stamford office fell
short of the goal in Quarter 1, but it has exceeded the goal in every subsequent quarter
and its booked revenues have been steady. Booked revenues have fallen short of the goal
in each quarter for the Providence office, but its booked revenues have been steadily
increasing. It is much more challenging to see these trends in Figure 3.28 because it is
more difficult to group the booked revenues for a location together without taking advan-
tage of the principle of connectedness. Figure 3.29 is also preferable because it reduces
the eye travel required of the audience through the removal of the legend from Figure
3.28 and the labeling of each line.

Trying to Display Too Much Information

Another common mistake in building effective charts is trying to convey too much infor-
mation on a single chart, which is a symptom of trying to communicate too many insights
to the audience simultaneously.
Consider the case of Keeland Industries, an online company that provides replace-
ment parts for automobiles. It provides both original-equipment manufacturer (OEM)
replacement parts and replacement parts made by different manufacturers that are
known as aftermarket replacement parts. Because Keeland sells the replacement parts
online, it sells parts to customers throughout the United States. For sales tracking
and performance measurement purposes, Keeland divides the United States into 12
regions. Keeland’s management team is most interested in comparing the OEM sales
across the 12 regions to see which are performing best for OEM sales and in compar-
ing the Aftermarket sales across the 12 regions to see which are performing best for
Aftermarket sales.
Figure 3.30 displays the OEM and Aftermarket sales by region as a clustered column
chart. The clustered column chart makes it easy to compare OEM sales to Aftermarket

FIGURE 3.30 Displaying Sales by Region for Keeland Industries as a

Clustered Column Chart

Keeland Sales Comparison

Sales ($1000s)
140 OEM
Aftermarket
120

100

80

60

40

20

0
1 2 3 4 5 6 7 8 9 10 11 12
Region
3-5 Common Mistakes in Data Visualization Design 105

sales in each region. However, if the potential markets for these different types of parts are
different in each region, that comparison is not particularly useful.
Figure 3.31 displays the OEM and Aftermarket sales by region as a stacked column
chart. The stacked column chart makes it much easier to compare the total sales in the
different regions. Therefore, a stacked column chart would be a good choice if the goal is
to compare total sales (OEM plus Aftermarket) among the 12 regions. However, the stated
goal of this visualization is to compare the OEM sales across regions separately from the
Aftermarket sales. The best visualization to accomplish the stated goal is to use two sepa-
rate column charts as shown in Figures 3.32 and 3.33.

FIGURE 3.31 Displaying Sales by Region for Keeland Industries as a

Stacked Column Chart

Keeland Sales Comparison

Sales ($1000s)

Aftermarket
200 OEM

150

100

50

0
1 2 3 4 5 6 7 8 9 10 11 12
Region

FIGURE 3.32 OEM Sales by Region for Keeland Industries

Keeland OEM Sales

Sales ($1000s)
140

120

100

80

60

40

20

0
1 2 3 4 5 6 7 8 9 10 11 12
Region
106 Chapter 3 Data Visualization and Design

FIGURE 3.33 Aftermarket Sales by Region for Keeland Industries

Keeland Aftermarket Sales

Sales ($1000s)
100

80

60

40

20

0
1 2 3 4 5 6 7 8 9 10 11 12
Region

Using Excel Default Settings for Charts

Microsoft Excel allows for the creation of a variety of charts and tables to visualize
data. However, a common mistake is to use the default output from Excel without
considering changes to the design and format of the visualizations it produces. Excel’s
default settings are counter to many of the suggestions covered in this chapter (and
the rest of this textbook) for creating good data visualizations. Consider Figure 3.34.
This column chart, which was produced using Excel, shows revenues for eight retail
store locations in Texas. The company is interested in comparing revenues by location,
and specifically in examining the relative performance of the store located in Laredo
because this store has recently had a change in management.

FIGURE 3.34 Default Column Chart Produced in Excel Comparing Annual

Revenue for Eight Retail Locations

Annual Revenue ($1000s)

400

300

200

Annual Revenue ($1000s)

100

0
Houston San Waco Dallas Amarillo Laredo El Paso Tyler
Antonio
Retail Location
 3-5 Common Mistakes in Data Visualization Design 107

Figure 3.34 suffers from several flaws that prevent it from being an effective data
visualization. The data-ink ratio for Figure 3.34 is low, so we should consider ways of
decluttering the figure. Examining Figure 3.34 shows that the chart uses ink in several
ways that are not useful in conveying the data. The gridlines used in this chart are not
particularly useful, so they can be removed. We see that Excel automatically titles the
chart “Annual Revenue” and uses a legend with “Annual Revenue.” This is redundant
information, and at least one of these labels should be removed. The following steps
can be used to declutter the default chart produced by Excel, increase the data-ink
ratio, and make the chart more meaningful to the audience.
Step 1. Click anywhere on the chart in the file RetailRevenueChart
Step 2. Click the Chart Elements button
Deselect the check box for Gridlines
RetailRevenueChart
Deselect the check box for Legend
Steps 1 and 2 increase the data-ink ratio by decluttering the chart. We can further
improve this chart by adding meaningful ink to the chart and making a few other mod-
ifications. For example, the revenue values shown in this chart are from the previous
year and are in 1000s of dollars. None of this is clear from the chart. To make it easier
for the audience to compare the relative amounts of annual revenue by location, we
can sort the columns in decreasing order. Finally, because the audience is particularly
interested in annual sales at the Laredo location, we can change the color of the col-
umn associated with Laredo to draw the audience’s attention to that part of the chart.
The following steps create the finished column chart shown in Figure 3.35.
Step 3. Select cells A1:B9
Click the Data tab on the Ribbon
Click Sort in the Sort & Filter group
Step 4. When the Sort dialog box appears:
Select the check box for My data has headers
In the Sort by box select Annual Revenue and in the Order box select
Largest to Smallest
Click OK
Step 5. Click the Chart Elements button
Select the check box for Data Labels
Step 6. Click on the vertical axis labels and press the Delete key
Step 7. Double-click the second column in the chart corresponding to “Laredo” so
that only this column is selected in the chart
In the Format Data Point task pane, click the Fill & Line icon
Under Fill, select a darker blue color for this column
Step 8. Click the chart title “Annual Revenue ($1000s)” and change this title to
Comparing Previous Year Annual Revenue at Laredo to Other Locations
Drag this text box to the left so that it aligns with the vertical axis
Click the Home tab on the Ribbon and select the Align Left button in
the Alignment group to left justify the chart title
Step 9. Click the Insert tab on the Ribbon
A

Click Text Box Text in the Text group and click above the vertical axis to
Box

create a text box for the vertical axis title

Type Revenue ($1000s) in the text box and change the font to Calibri
10.5 Bold to create the vertical axis title (see Figure 3.35)
108 Chapter 3 Data Visualization and Design

FIGURE 3.35 Completed Column Chart Comparing Annual Revenue for

Eight Retail Locations

Comparing Previous Year Annual Revenue at Laredo to Other Locations

Revenue ($1000s)
350
325

195 185
150

85 75

25

Dallas Laredo Waco Houston El Paso San Antonio Amarillo Tyler

Retail Location

Steps 3 and 4 sort the columns to make relative comparisons of the revenue by location
using the attribute of length more obvious to the audience. Step 5 adds data labels at the
top of each column to give the exact value of revenues at that location. This should only be
done when the audience may need to know the exact values shown in the chart. Because
we have added data labels to each column, we remove the redundant vertical axis labels
in Step 6. Step 7 uses the preattentive attribute of color to differentiate the column corre-
sponding to Laredo to draw the audience’s attention to this column. Finally, Steps 8 and 9
align the chart title and vertical-axis title with the vertical axis to minimize the required eye
travel to interpret this chart.

Too Many Attributes

In Section 3-1, we discuss the importance of using preattentive attributes in data visu-
alizations to make them easy to understand by the audience. However, using too many
preattentive attributes in the same visualization can cause confusion for the audience.
Consider again the case of Stanley Consulting Group. The company wants to examine
how consultant characteristics such as job title, length of time with the company, and
highest educational degree attained are related to the amount of billable hours filed by
that consultant. Figure 3.36 attempts to show this information.
All of the information the company wants to consider is shown in Figure 3.36: the num-
ber of billable hours for each consultant (on the vertical axis), the length of time at the com-
pany (on the horizontal axis), the consultant’s job title (indicated by the color of the marker
in the chart), and the highest degree attained by the consultant (indicated by the shape of
the marker in the chart). Figure 3.36 uses several preattentive attributes from Section 3-1
including spatial positioning, shape, and color. However, because we are using many dif-
ferent preattentive attributes, this chart is difficult for an audience to process. It requires the
Figure 3.36 can be an effective audience to scan back and forth between the markers in the chart, the legends, and the verti-
chart for the analyst who is cal and horizontal axes. Therefore, this is probably not a particularly useful chart.
exploring the data looking for A better chart than what is shown in Figure 3.36 would concentrate on examining fewer
relationships. However, it is
relationships and using fewer preattentive attributes. The exact choice of which features to
probably not the most effective
choice for presenting to an
show on the chart depends on the goals of the chart and needs of the audience. If it is more
audience as a final version important to examine the relationship between billable hours, length of time at the com-
meant to explain the data. pany, and the job title of the consultant, then a chart such as the one shown in Figure 3.37 is
preferred.
 3-5 Common Mistakes in Data Visualization Design 109

The type of marker displayed

on a chart in Excel can be FIGURE 3.36 Chart Displaying the Association of Billable Hours with Years
changed in the Format Data at Company, Job Title, and Highest Degree Attained
Series task pane by clicking at Stanley Consulting Group
on the Fill & Line icon ,
clicking on Marker, then
Marker Options and selecting Billable Hours by Consultant Job Title (Color) Highest Degree (Shape)
a marker shape from Built-in.
Hours (per month) Junior Consultant Bachelor
Senior Consultant Master
200 Partner PhD

150

100

50

0
0 2 4 6 8 10 12 14 16 18
Years at Company

FIGURE 3.37 Simplified Chart Showing the Association of Billable Hours with
Years at Company and Job Title for Stanley Consulting Group

Billable Hours by Consultant Junior Consultant

Hours (per month)
Senior Consultant
200 Partner

150

100

50

0
0 2 4 6 8 10 12 14 16 18
Years at Company

Unnecessary Use of 3D
Three-dimensional (3D) charts are often difficult for an audience to understand. Many nov-
ice analysts will create 3D charts in Excel even when the third dimension does not provide
any useful information. Consider the 3D column chart shown in Figure 3.38. This chart
compares the billable hours for eight consultants who work for Stanley Consulting Group.
110 Chapter 3 Data Visualization and Design

FIGURE 3.38 a 3D Column Chart Displaying Billable Hours for Stanley

Group Consultants Where the Third Dimension Provides No
Useful Information

Billable Hours By Consultant

160
140
120
Hours (per month)

Marilyn
100
Jing
80 Glenn
60 Rasesh
Larry
40 Aubree
20 Bill
Ashwita
0

Because the third dimension in Figure 3.38 is not used to display any useful information, it
only serves to decrease the data-ink ratio of the chart and make it more difficult for the audi-
ence to interpret. Even when the third dimension is used to display unique information, 3D
charts are difficult to interpret. Therefore, we generally recommend against the use of 3D
charts for data visualization. Therefore, in a better design for this column chart, we remove
the third dimension and display the chart as the 2D column chart as shown in Figure 3.39.

FIGURE 3.39 An Improved 2D Column Chart Displaying Billable Hours for

Stanley Group Consultants

Billable Hours by Consultant

Hours (per month)
160

140

120

100

80

60

40

20

0
Ashwita Bill Aubree Larry Rasesh Glenn Jing Marilyn
Consultant
Glossary 111

S u mmar y

In this chapter, we have discussed how to use specific design elements to create effec-
tive data visualizations. We introduced the preattentive attributes of color, form, spatial
positioning, and movement. We showed how these attributes can be used to make data
visualizations easier to interpret for audiences and how they can direct the audience’s
attention. We also introduced the Gestalt principles of similarity, proximity, enclosure,
and connection as they relate to how people interpret objects in a data visualization. The
proper use of preattentive attributes and Gestalt principles can help create effective data
visualizations. We then discussed strategies for minimizing eye travel for the audience,
and choosing an effective font for text in visualizations to make charts and tables as easy
to interpret as possible. We applied these design concepts to tables, scatter charts, bar
charts, column charts, and line charts. Finally, we covered some common mistakes in data
visualization design such as choosing the wrong type of visualization, trying to display
too much information in a single visualization, using the default format for charts created
in Excel, and unnecessarily using 3D charts. Throughout the chapter, we emphasized that
what makes a good data visualization depends on the needs of the audience and the goal
of the visualization. In later chapters, we will cover these issues in additional detail.

G L O SS A RY

Cognitive load The amount of effort necessary to accurately and efficiently process the
information being communicated by a data visualization.
Color A preattentive attribute for data visualizations that includes the attributes of hue,
saturation, and luminance.
Connection A Gestalt principle stating that people interpret objects that are connected in
some way as belonging to the same group.
Data-ink ratio Measures the proportion of “data-ink” to the total amount of ink used in a
table or chart where data-ink is the ink used that is necessary to convey the meaning of the
data to the audience.
Decluttering The act of removing the non-data-ink in the visualization that does not
help the audience interpret the chart or table.
Enclosure A Gestalt principle stating that objects that are physically enclosed together are
seen as belonging to the same group.
Flicker A movement type of preattentive attribute that refers to effects such as flashing to
draw attention to something.
Form A collection of the preattentive attributes of orientation, size, shape, length, and width.
Gestalt principles The guiding principles of how people interpret and perceive things that
they see, which can be used in the design of effective data visualizations. The principles
generally describe how people define order and meaning in things that they see.
Hue the attribute of a color that is determined by the position the light occupies on the
visible light spectrum and defines the base of the color.
Iconic memory The portion of memory that is processed fastest. It is processed
automatically and the information is held there for less than a second.
Length A type of preattentive attribute associated with form. It refers to the horizontal,
vertical, or diagonal distance of a line.
Long-term memory The portion of memory where information is stored for an extended
amount of time. Most long-term memories are formed through repetition and rehearsal but
can also be formed through clever use of storytelling.
Luminance The attribute of a color that represents the relative degree of black or white in
the color.
Motion A movement type of preattentive attribute that involves directed movement and
can be used to show changes within a visualization.
Orientation A type of preattentive attribute associated with form. It refers to the relative
positioning of an object within a data visualization.
112 Chapter 3 Data Visualization and Design

Preattentive attributes Features of a data visualization that can be processed by iconic

memory. Preattentive attributes related to visual perception are generally divided into four
categories: color, form, spatial positioning, and movement.
Proximity A Gestalt principle stating that people consider objects that are physically close
to one another as belonging to a group.
Sankey chart A type of data visualization chart that typically depicts the proportional
flow of entities where the width of the line represents the relative flow rate compared to the
widths of the other lines.
Sans-serif font A style of text that does not include small end-of-stroke features on the
characters. These types of fonts include Arial, Calibri, Myriad Pro and Verdana and are
generally preferred for text in data visualizations.
Saturation The attribute of a color that represents the amount of gray in the color and
determines the intensity or purity of the hue in the color.
Serif font A style of text that includes small end-of-stroke features on the characters.
These types of fonts include Times, Times New Roman, and Courier.
Serifs The small end-of-stroke features that are visual in the characters created using serif fonts.
Shape A type of preattentive attribute associated with form. It refers to the type of object
used in a data visualization.
Short-term memory The portion of memory that holds information for about a minute.
It utilizes chunking, or grouping things together to hold about four chunks of visual
information at one time.
Similarity A Gestalt principle stating that people consider objects with similar
characteristics as belonging to the same group.
Size A type of preattentive attribute associated with form. It refers to the relative amount
of 2D space that an object occupies in a visualization.
Spatial positioning A preattentive attribute that refers to the location of an object within
some defined space.
Visual perception The process through which our brains interpret the reflections of light
that enter our eyes.
White space The portion of a data visualization that is devoid of markings.
Width A type of preattentive attribute associated with form. It refers to the thickness of
the line.

P ro b l e ms

CONCEPTUAL

1. Memory Used for Preattentive Attributes. Which of the following types of memory
is used to process preattentive attributes? LO 1
i. Iconic memory
ii. Short-term memory
iii. Long-term memory
iv. Random access memory
2. Preattentive Attributes in a Data Visualization. Which of the following statements
about the use of preattentive attributes in a data visualization are true? (Select all that
apply.) LO 1
i. The use of preattentive attributes reduces the cognitive load required by the audi-
ence to interpret the information conveyed by a data visualization.
ii. Preattentive attributes can be used to draw the audience’s attention to certain parts
of a data visualization.
Problems 113

iii. Overuse of preattentive attributes can lead to clutter and can be distracting to the
audience.
iv. Preattentive attributes include attributes such as proximity and enclosure.
3. Descriptions of Gestalt Principles. For each description below, provide the name of
the Gestalt principle that is being described. LO 2
a. Objects that are physically close to one another are seen as belonging to the same
group.
b. Objects that are linked in some way are seen as belonging to the same group.
c. Objects that are physically bound together are seen as belonging to the same group.
d. Objects with like characteristics such as color, shape, size, etc. are seen as belong-
ing to the same group.
4. Scatter with Straight Lines and Markers Chart in Excel. Using a Scatter with Straight
Lines and Markers Chart in Excel makes use of which Gestalt principle? LO 2
i. Similarity
ii. Proximity
iii. Enclosure
iv. Connection
5. Increasing the Data-Ink Ratio on a Chart. Which of the following changes to a chart
would increase the data-ink ratio? (Select all that apply.) LO 3
i. Removing unnecessary gridlines.
ii. Removing a legend on a bar chart where each bar is already labeled with the
same information.
iii. Adding axes labels to a chart where the units used in each axis are not clear from
the chart title.
iv. Adding data labels for each point on a scatter chart when the audience does not
need to know exact values.
6. Pie Charts versus Bar or Column Charts. Which of the following reasons accurately
describe why bar or column charts are often preferred to pie charts for a data visualiza-
tion? (Select all that apply.) LO 4
i. Bar and column charts utilize the Gestalt principle of proximity while pie charts
use the Gestalt principle of connection.
ii. Using a legend for a pie chart creates unnecessary eye travel that can often be
reduced by using a bar or column chart that does not require a legend.
iii. Bar and column charts use length rather than size to make comparisons, and
length is much easier for the audience to interpret than size.
iv. Pie charts often use different colors to differentiate each piece of the pie that can
create unnecessary clutter compared to bar or column charts that can display the
same information without the use of multiple colors.
7. Market Segmentation for Brandience. Brandience Marketing LLC provides mar-
keting analytics consulting for clients. For one of its clients, Brandience has been
asked to perform a market segmentation study for a business client that provides
auditing services to manufacturing customers. The client believes there are two vari-
ables of importance that should be used to group similar customers into clusters:
Years of Service with the Client and Total Assets. Brandience plans to use a cluster-
ing algorithm to group similar customers into different clusters, but before applying
the algorithm, Brandience creates a simple scatter chart to plot each customer based
on their Years of Service with the Client and Total Assets. The scatter chart created
by Brandience follows. LO 2
114 Chapter 3 Data Visualization and Design

Market Segmentation Analysis of Brandience Clients

Total Assets ($ millions)
30

25

20

15

BrandienceChart
10

0
0 2 4 6 8 10 12
Years of Service

a. Based on the Gestalt principle of proximity, how many different groups of custom-
ers are shown in the scatter chart?
b. What other Gestalt principle could be used to reinforce the appearance of the differ-
ent groups of customers for the audience?
8. Philadelphia City Schools. The following chart displays data related to the student
enrollment at several schools located in the Philadelphia City School District. LO 1, 4

Northeast High Central High Shawmont

John F. McCloskey George Washington High William H. Loesche
Lankenau High Anne Frank Stephen Decauter

a. Which preattentive attributes are being used in this chart?

b. What changes would you suggest to this chart to make it easier for the audience to
interpret?
9. Percentage of College Graduates and Median Monthly Rent in New York City.
The following scatter chart shows the relationship between the percentage of college
graduates and the median monthly rent in different sub-boroughs (a description of
an area within the city) of New York City. Each point in the scatter chart represents a
different sub-borough in New York City. LO 1
Problems 115

College Graduates (%)

100

80

60

40

20

0
0 500 1000 1500 2000 2500
Median Monthly Rent ($)

a. Which preattentive attribute is used in this scatter chart?

b. What can you say about the relationship between the percentage of college gradu-
ates and the median monthly rent in sub-boroughs of New York City?
10. Patient Waiting Times at Scott & White. The Scott & White Medical Center in Temple,
Texas, is examining the mean waiting time for patients to receive common orthopedic
surgeries at its facility. The medical administrators have collected data on patient waiting
times. The mean patient waiting times (in days) for different orthopedic surgeries are
shown in the chart below. The medical center is particularly interested in the wait times
for anterior cruciate ligament (ACL) reconstruction surgeries because ACL reconstruc-
tions are the most common type of orthopedic surgery performed at its facility. LO 1

Comparing Patient Waiting Times at Scott & White Medical Center

Surgery Type

Ankle Repair

Hip Replacement

Joint Fusion

Knee Replacement

Shoulder Replacement

ACL Reconstruction

Knee Arthroscopy

Shoulder Arthroscopy

0 5 10 15 20 25 30 35 40 45 50
Average Waiting Time (days)

What types of preattentive attributes are used in this chart and what information do
these attributes convey to the audience?
116 Chapter 3 Data Visualization and Design

11. Comparing Managing Consultants at Platt Consulting. Platt Consulting Services is

interested in comparing the number of customer accounts managed by its consultants.
In particular, the company needs to compare the number of accounts managed by
Bernie Smith, Stanley Lucas, and Gracie Rogers because these three accountants
started with the company at approximately the same time and have similar skills.
The chart below is a default chart produced in Excel. LO 1
Number of Customer Accounts Managed

Norma Lane Caroline Hyde Margaret Walnut Michael Withrow Hurley Reyes
Stanley Lucas Gracie Rogers Bernie Smith Kendall Espinosa

a. Which preattentive attributes are used in the pie chart?

b. What preattentive attributes would you use in redesigning this chart to improve its
clarity for the audience?
12. Basketball Scouting Report. Meg Lawson is an assistant coach for a high school basket-
ball team. Meg has prepared a scouting report on her team’s upcoming opponent, which
is led by a star player who prefers to shoot from near the 3-point line. Meg has created the
following diagram to show this star player’s shot selection from a recent game. Each ×
corresponds to a missed shot and each • corresponds to a made shot. Meg has noticed that
the star player made most of her shots from the area on the court located at the top of the
3-point line while missing most of her shots from the corners. Meg wants to draw her team’s
attention to these areas so that they can properly defend against this star player. LO 1, 2
Problems 117

a. What preattentive attributes and Gestalt principles has Meg used in this diagram?
b. Are the preattentive attributes and Gestalt principles used in this diagram effective
in communicating Meg’s message to her audience? Why or why not?
13. Revenues and Costs at a Manufacturing Firm. A manufacturing firm would like
to compare its monthly revenues and costs for the previous year. The following table
shows the revenues and costs for the company in each month of the previous year. Of
particular interest to the company is how the company’s revenues and costs perform
during the summer months because that is when the company’s sales tend to be
highest. LO 3

Month Revenues ($1000s) Costs ($1000s)

Jan 1121 1007
Feb 997 1002
Mar 1151 1010
Apr 1202 1085
May 1422 1287
Jun 1877 1488
Jul 1911 1621
Aug 1988 1625
Sep 1521 1617
Oct 1288 1178
Nov 1100 1008
Dec 1022 987

Previous Year Revenues and Costs

2500

2000

1500
Amount

1000

500

0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
Revenues Costs
Chart A
118 Chapter 3 Data Visualization and Design

Previous Year Revenues and Costs

($1000s)

2500

2000

1500

Revenues
1000
Costs
500

Summer
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Chart B

a. Compare the two charts above (Chart A and Chart B). Which chart has the higher
data-ink ratio? Why?
b. How does the chart with the higher data-ink ratio help the audience interpret the
chart more effectively?
14. Serif versus Sans-Serif Fonts. Consider the fonts that are shown in the table that
follows. LO 5

(1) A b c d 1 2 3 (4) A b c d 1 2 3 (7) a b c d 1 2 3

(2) A b c d 1 2 3 (5) A b c d 1 2 3 (8) A b c d 1 2 3
(3) A b c d 1 2 3 (6) A b c d 1 2 3 (9) A b c d 1 2 3
a. Which of these fonts are considered serif fonts?
b. If you are creating text for a data visualization that will mostly be viewed on smart-
phones and other small mobile devices, would you suggest using the serif fonts for
text in the data visualization? Why or why not?
15. Approval Voting Results. Approval voting is a type of voting system in which voters
can vote for any number of eligible candidates. A vote for a candidate in this type of
system indicates that the voter “approves” of that candidate for the position the can-
didate is seeking. In the final tally of the election results, the total number of approval
votes for each candidate is calculated. The candidate who receives the most approval
votes is declared the winner. The figure below displays the results from an approval
voting election in which there were four eligible candidates: C. Sittenfeld, A. Marshall,
S. Keskin, and K. Nowak. A total of 1218 voters participated in this election. The table
below shows the number of votes received by each candidate as well as the proportion
of voters who approved of each candidate. LO 4, 6

Proportion of Voters Who

Candidate Number of Votes Received Approve of This Candidate
C. Sittenfeld 482 39.6%
A. Marshall 689 56.6%
S. Keskin 354 29.1%
K. Nowak 514 42.2%
Problems 119

Proportion of Voters Who Approve of Candidate

42.2%
39.6%

29.1%
56.6%

C. Sittenfeld A. Marshall S. Keskin K. Nowak

a. The figure displays the election result proportions as a 3D pie chart. What problems
are associated with displaying the election results in this way?
b. Suggest an alternative type of chart for displaying these data that would be easier
for the audience to interpret. Explain why this alternative type of chart would be
easier to interpret.
16. Headcount and Annual Revenues at BeFit Gyms. BeFit Gyms operates four exercise
gyms in the state of Michigan. The gyms are located in Saline, Tecumseh, Dexter, and
Jackson. The gyms are different sizes, and they require different headcounts of staff
to operate. The table below displays the headcount, in full-time equivalent (FTE)
employees and annual revenue ($1000s) for each of the four locations. LO 4, 6

Gym Location Headcount (FTE) Annual Revenue ($1000s)

Saline 18 787
Tecumseh 11 674
Dexter 9 784
Jackson 12 642

Comparing Headcount and Annual Revenue Across Locations

900

800

700

600

500

400

300

200

100

0
Saline Tecumseh Dexter Jackson
Headcount (FTE) Annual Revenue ($1000s)
120 Chapter 3 Data Visualization and Design

a. The clustered column chart displays the headcount and annual revenue for each
location. What are the problems with displaying these data using this type of
chart?
b. Would it be appropriate to display these data using a stacked column chart rather
than a clustered column chart?

APPLICATIONS

17. Comparing Managing Consultants at Platt Consulting (revisited). In this prob-

lem, we revisit the pie chart in Problem 11. Platt Consulting Services is interested in
comparing the number of customer accounts managed by its managing consultants. In
PlattConsulting particular, the company needs to compare the number of accounts managed by Bernie
Smith, Stanley Lucas, and Gracie Rogers because these three accountants started with
the company at about the same time and have similar skill sets. The following chart is a
default chart produced in Microsoft Excel. LO 1, 6
Number of Customer Accounts Managed

Norma Lane Caroline Hyde Margaret Walnut Michael Withrow Hurley Reyes
Stanley Lucas Gracie Rogers Bernie Smith Kendall Espinosa

Use the data in the file PlattConsulting to create a new data visualization that uses a
different type of chart and proper use of preattentive attributes to allow for easier com-
parison of the number of accounts managed by each consultant.
18. Market Segmentation for Brandience (revisited). In this problem, we revisit the
scatter chart in Problem 7. Starting with the chart in the file BrandienceChart, modify
the chart by using an additional Gestalt principle to make it more obvious to the audi-
BrandienceChart
ence which clients are in each cluster. Which additional Gestalt principle did you use?
LO 2, 4

19. Comparing Direct and Indirect Costs at Sackenheim Compressors. Sackenheim

Compressors, Inc. manufactures air compressors for industrial use. The company is
performing an analysis of its direct and indirect costs related to manufacturing a par-
ticular type of air compressor. It has gathered cost data for the previous 20 weeks, and
it would like to identify any trends in the data. A scatter chart created using the data
follows. LO 2, 4
 Problems 121

Direct vs. Indirect Cost Comparison

Costs ($1000s)
18 Direct Indirect
16
14
12
10
8
SackenheimChart
6
4
2
0
0 2 4 6 8 10 12 14 16 18 20
Week

a. Which Gestalt principle should be used here to make it easier for the audience to
identify trends in the data?
b. Using the chart included in the file SackenheimChart as a starting point, create an
improved version of this chart by applying the Gestalt principle from part a that
makes any trends in the data more obvious to the audience.
20. Sales Performance Bonuses. A sales manager is trying to determine appropriate sales
performance bonuses for her team this year. The following table contains the data rel-
evant to determining the bonuses, but it is not easy to read and interpret. Reformat the
table using the data in file SalesBonuses to improve readability and to help the sales
manager make her decisions about bonuses. (Hint: It will also help the sales manager if
the table is ordered from top-to-bottom by Total Sales.) LO 3

Average ­Performance
Total Sales Bonus Previous Customer Years with
Salesperson ($) Years ($) Accounts Company

Smith, Michael 325000.78 12499.3452 124 14

Yu, Joe 13678.21 239.9434 9 7
Reeves, Bill 452359.19 21987.2462 175 21
Hamilton, Joshua 87423.91 7642.9011 28 3
Harper, Derek 87654.21 1250.1393 21 4
Quinn, Dorothy 234091.39 14567.9833 48 9
SalesBonuses
Graves, Lorrie 379401.94 27981.4432 121 12
Sun, Yi 31733.59 672.9111 7 1
Thompson, Nicole 127845.22 13322.9713 17 3
122 Chapter 3 Data Visualization and Design

21. Gross Domestic Product Values over Time. The following table shows an example of
gross domestic product values for five countries over six years in equivalent U.S. dollars ($).
LO 2, 3

Gross Domestic Product (US $)

Country Year 1 Year 2 Year 3 Year 4 Year 5 Year 6

Albania 7385937423 8105580293 9650128750 11592303225 10781921975 10569204154
Argentina 169725491092 198012474920 241037555661 301259040110 285070994754 339604450702
Australia 704453444387 758320889024 916931817944 982991358955 934168969952 1178776680167
Austria 272865358404 290682488352 336840690493 375777347214 344514388622 341440991770
Belgium 335571307765 355372712266 408482592257 451663134614 421433351959 416534140346

a. How could you improve the readability of this table?

b. The file GDPYears contains sample data from Year 1 to Year 6 in us $. Create a
table that provides all these data for an audience. Format the table to make it as
GDPYears
easy to read as possible. Hint: It is generally not important for the audience to know
GDP to an exact dollar figure. It is typical to present GDP values in millions or
billions of dollars.
22. Monthly Revenue Data at Tedstar, Inc. The following table provides monthly reve-
nue values for Tedstar, Inc., a company that sells valves to large industrial firms. The
monthly revenue data have been graphed using a line chart in the following figure.
LO 3, 4

Month Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Revenue ($) 145,869 123,576 143,298 178,505 186,850 192,850 134,500 145,286 154,285 148,523 139,600 148,235

210000
200000
190000
180000
170000
160000
150000
140000
Revenue ($)

130000
120000
110000
100000
90000
80000
70000
Tedstar 60000
50000
40000
30000
20000
10000
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Months

a. What characteristics of this line chart make it more difficult than necessary to
interpret?
b. Using the data in the file Tedstar, create a new line chart for the monthly reve-
nue data at Tedstar, Inc. Format the chart to make it easy to read and interpret.
Use the chart title “Tedstar, Inc. Revenue Analysis” and the vertical-axis title
“Monthly Revenue ($).”
23. Comparing Hoxworth Blood Center Locations. Hoxworth Blood Center, located in
Cincinnati, Ohio, is a leader in transfusion medicine. Founded in 1938, it is the second
oldest blood bank in the United States. Hoxworth operates at seven locations in the
Greater Cincinnati area where it collects blood from donors. Hoxworth is performing
HoxworthChart a comparison of the amount of blood donors serviced by each of its locations during
 Problems 123

the month of October. Suppose that Hoxworth has a stated goal of each location servic-
ing an average of 50 blood donors per day each month. The column chart that follows
compares the average number of blood donors at each Hoxworth location. However,
this chart is cluttered and the data-ink ratio is low. Starting with the chart in the file
HoxworthChart, declutter the chart and improve the data-ink ratio to produce an
improved column chart. LO 3, 4
Hoxworth Comparison of Number of Donors Serviced by Location
70

Average Number of Donors Serviced 60

Goal = Average
per Day During October
50
50 Donors per day
40

30

20 Number of Donors
Serviced

10

0
Anderson Blue Ash Central Fort Mitchell North Tri-County West
Locations

24. Comparing Package Delivery Drivers. Red Sky Delivery performs “last-mile” deliv-
ery services for online retailers such as Amazon. Red Sky employs delivery drivers
who perform the “last-mile” of delivery service by delivering packages to individual
residence and business locations. Red Sky measures several delivery driver perfor-
mance metrics, including number of delivery stops completed per eight-hour shift. The
table below provides data on nine Red Sky delivery drivers and their average number
of packages delivered per shift over the previous 30 days. LO 1, 3, 4

Average Number of
Delivery Stops Completed
Delivery Driver (per shift)
Amy Urbaczewski 92.87
Sally Melouk 110.78
Brenda Barnes 114.20
Jonathan Payne 132.50
RedSkyDeliveries
Bruce Wheeler 148.20
Cam Madsen 87.51
Sheila Stevens 139.82
Grant Inhara 154.23
Finn Helton 109.11

a. Using the data in the file RedSkyDeliveries, create a column chart to display the infor-
mation in the table above. Format the column chart to best display the data. Use
a chart title of “Comparing Red Sky Delivery Drivers” and a vertical-axis title of
“Average number of deliveries per shift.” Get rid of any unnecessary gridlines and add
data labels that show the average number of delivery stops completed for each driver.
124 Chapter 3 Data Visualization and Design

b. Create a sorted column chart to make it easier for the audience to see which drivers
have the highest and lowest average number of delivery stops per shift.
c. Further investigation by Red Sky indicates that all of these delivery drivers
except Amy Urbaczewski have similar delivery routes. Amy typically deliv-
ers in more rural areas while all other drivers support more urban routes. Red
Sky wants to draw attention to the fact that Amy’s routes are different than the
others. Modify the sorted column chart by changing the color of the column
associated with Amy Urbaczewski to indicate that this column is different from
the others.
25. Inbound and Outbound Shipping Costs at Rainbow Camping. Rainbow Camping
makes outdoor equipment for camping and backpacking that it sells online. Because
Rainbow Camping mails each of its sales from its distribution center direct to the
consumer, it spends considerable funds each month on outbound shipping. Rain-
bow Camping also has costs related to inbound shipping to get its products from the
manufacturers to its distribution center. Rainbow Camping would like to examine its
outbound and inbound shipping expenses over the last 12 months (labeled months 1
through 12). These shipping costs are shown in the following table. LO 4

Shipping Costs ($1000s)

Month Outbound Inbound
1 478 324
2 524 274
3 628 224
4 787 292
5 648 348
6 612 309
RainbowCamping
7 598 378
8 641 287
9 546 292
10 476 345
11 378 304
12 512 298

a. Using the data in the file RainbowCamping, a single chart that displays both the
outbound and inbound shipping costs for each month on the same chart. Use
the title “Rainbow Camping Shipping Costs Analysis” for the chart title, “Costs
($1000s)” for the vertical-axis title, and “Month” for the horizontal-axis title.
Format the chart title and vertical-axis title to make these easy for the audience to
interpret. Modify the Minimum and Maximum Bounds of the horizontal axis so that
this axis starts at 1 and ends at 12 (Minimum and Maximum Bounds can be found
in the Format Axis task pane by clicking on the Axis Options icon and then
under Axis Options). Remove the markers so that only lines are displayed for each
of these shipping costs. Delete any unnecessary gridlines in the chart.
b. To further minimize the eye travel required by the audience, delete the default
legend created with the chart and use text boxes to label the lines in the chart as
“Outbound” and “Inbound.”
26. University of Michigan Enrollments over Time. The University of Michigan was
founded in 1817 as the first university in what was then the Northwest Territories of
the United States. The university’s nickname is the Wolverines and its main campus is
MichiganEnrollment
located in Ann Arbor, Michigan. The file MichiganEnrollment contains enrollment data
for students at the University of Michigan’s Ann Arbor campus for Fall Semesters from
 Problems 125

1966 to 2016 in five-year increments. Create a default chart for these data by clicking
Insert in the Ribbon in Excel and then choosing Scatter with Straight Lines and
Markers from the Charts group. Use the principles of data-ink ratio and declut-
tering to improve this chart for the audience. LO 3, 4
In particular, you should:
●● Create meaningful chart and vertical axis title, and position them appropriately to

minimize eye travel.

●● Adjust the Minimum and Maximum Bounds of the horizontal axis to make it clear

to the audience that the data points correspond to 1966, 1971, 1976, etc. (Minimum
and Maximum Bounds can be found in the Format Axis task pane by clicking on
the Axis Options icon and then under Axis Options).
●● Create an appropriate horizontal-axis title to ensure that the audience understands

that these values correspond to Fall Semester enrollments and position this axis title
to minimize eye travel.
●● Remove the gridlines and adjust the vertical-axis units to increase the data-ink ratio.

●● Make any other improvements to the chart that you think would help the audience

interpret this chart.

27. Evaluating Marketing Strategies. Cuero Marketing Services (CMS) is a marketing
agency that designs marketing plans for companies. CMS is working with a large global
corporation on new brand strategies. CMS formed seven different test groups to evaluate
how people perceive two competing ideas for new brand strategies. Each test group is
composed of 25 people. CMS collects all data from the test groups, and for each mem-
ber of the test group it creates an overall evaluation of whether the test group member
has an overall good impression of the brand strategy or does not have an overall good
CMS impression of the brand strategy. To preserve confidentiality, CMS refers to the two
competing brand strategies in all written documents as “Blue Triangle” and “Orange
Square.” The table below shows how many test group members from each test group
had an overall good impression of the Blue Triangle and Orange Square brand strate-
gies. CMS would like to create a simple data visualization to show these data. LO 4, 6

Number of Positive Impressions

Test Group Number Blue Triangle Orange Square
1 17 12
2 21 15
3 16 13
4 19 14
5 23 17
6 26 18
7 20 16

a. Using the data in the file CMS, create a scatter chart that shows both data series in
a single chart. Use the test group number on the horizontal axis and the number of
test group members who had an overall good impression on the vertical axis. Use
“Test Group Analysis for Brand Strategies” as the chart title, “Number of Members
with Overall Good Impression of Strategy (out of 25)” as the vertical-axis title, and
“Test Group” as the horizontal-axis title. Format the title to minimize eye travel and
remove any unnecessary gridlines.
b. To make the chart more intuitive for the audience, CMS would like to display
markers that match the name of each brand strategy. To do this, remove the default
legend from chart and replace the markers for the Blue Triangle brand strategy with
blue triangles and use orange squares for the markers of the Orange Square brand
strategy. (Hint: The markers on a chart can be can be changed in the Format Data
126 Chapter 3 Data Visualization and Design

Series task pane by clicking on the Fill & Line icon , clicking on Marker, then
Marker Options and selecting a marker shape from Built-in.)
c. Change the type of chart used to display these data to a clustered column chart. Do
you think the scatter chart or the clustered column chart would be most effective for
displaying these results to an audience? Why?
28. Affinity and Capacity of Donors for Malaria No More. Nonprofits often score
potential donors on multiple dimensions including affinity and capacity. Affinity
attempts to measure how passionate and engaged the potential donor is regarding the
MalariaChart
nonprofit’s cause. Capacity attempts to measure the donor’s available wealth and abil-
ity to donate to the nonprofit. Malaria No More is a nonprofit that focuses on eradicat-
ing malaria across the world. Suppose that Malaria No More measures the affinity and
capacity of potential donors on scales of 1–100 where 1 corresponds to little affinity
or capacity and 100 corresponds to extreme affinity or capacity. The file MalariaChart
contains a scatter chart showing a sample of 50 potential donors to Malaria No More
including their affinity and capacity scores. LO 2, 4
a. Starting with the scatter chart in the file MalariaChart, improve the design of this
chart to increase its effectiveness. Change the title of this chart to “Evaluating Po-
tential Donors for Malaria No More.” Add a vertical-axis title of “Capacity Score”
and a horizontal-axis title of “Affinity Score.” Format the chart title and axes titles
to minimize eye travel.
b. Malaria No More would like to improve this chart to convey a specific message to
its audience. It wants to focus the audience’s attention on those potential donors
with capacity and affinity scores greater than or equal to 80 because it believes that
these are the best donor prospects. Use the Gestalt principle of enclosure to high-
light the potential donors on the chart. (Hint: To create an enclosure in Excel, click
Insert on the Ribbon, then click Shapes in the Illustrations group.)
Shapes
29. Funstate Carnivals Staffing Analysis. Funstate Carnivals operates four large amuse-
ment parks, each of which is located in California. Funstate would like to examine the
staffing across these four amusement parks. It is most interested in comparing the total
number of staff employed at each park, but it is also interested in comparing the gender
breakdown of the staffing at each park. The table below shows the total number of staff
employed at each park, broken out by gender. LO 4

Total Number of Staff

Location Male Female
Fresno 72 84

Funstate
Sacramento 89 61
Long Beach 65 84
Oakland 48 51

a. Using the data in the file Funstate, create a stacked bar chart to display these data.
Use different colors for Male and Female. Use the chart title “Funstate Carnivals
Staffing Analysis,” choose appropriate axes titles, and include a chart legend to
identify Male versus Female in the bar chart. Format the chart to minimize eye
travel and remove any unnecessary gridlines to increase the data-ink ratio.
b. Funstate would like to show the exact numbers of Male and Female staff at each
location on the chart. Add data labels to the stacked bar chart to display the number
of Male and number of Female staff at each location. Format the data labels so they
are easy for the audience to read.
 Problems 127

c. Because of decreased revenues, Funstate wants to implement a policy of no more

than 125 total staff members at each location. Draw a line and use a text box
to indicate “Staffing Goal = 125” on the stacked bar chart to show this limit.

(Hint: Click Insert on the Ribbon, then click Shapes in the Illustrations group
Shapes

to add a line and text box.) Which locations currently exceed this staffing limit?
30. Approval Voting Results (revisited). In this problem, we revisit the data from Problem
15 on the results of an election using approval voting. The original figure in Problem 15
used a pie chart to display the results from an approval voting election in which there
ApprovalVoting
were four eligible candidates: C. Sittenfeld, R. Manley, S. Keskin, and K. Nowak. A
total of 1218 voters participated in this election. LO 3, 4
a. Using the data in the file ApprovalVoting, create a sorted bar chart that displays
the proportion of voters who approve of the candidate for each candidate. Choose
appropriate titles for the chart and axes. Sort the bars so that the candidate who
received the most votes is at the top. Use data labels to display the proportion of
votes received by each candidate. Format the chart title to minimize eye travel and
remove any unnecessary gridlines to increase the data-ink ratio. Which candidate
should be declared the winner of this election?
b. Create a sorted bar chart that displays the number of votes received by each can-
didate. Choose appropriate titles for the chart and axes. Sort the bars so that the
candidate who received the most votes is at the top. Use data labels to display the
number of votes received by each candidate. Format the chart title to minimize eye
travel and remove any unnecessary gridlines to increase the data-ink ratio. Which
candidate should be declared the winner of this election?
c. Do you think the sorted bar chart in part a or in part b is more effective for commu-
nicating the results of the election to an audience? Why?
31. Headcount and Annual Revenues at BeFit Gyms (revisited). In this problem,
we revisit the data from Problem 16 on staff headcount and annual revenues for
BeFit Gyms. Recall that each gym is a different size. To support larger headcounts,
larger gyms are expected to generate greater annual revenue. BeFit is interested in
BeFit
identifying the incremental revenue a location should expect from a larger head-
count. In problem 16, a clustered column chart was used to display these data.
However, we would like to make it easier for the audience to make comparisons
from these data and generate insights about the relation between headcount and
annual revenue. LO 3, 4, 6
a. Using the data in the file BeFit, create two different column charts to display the
headcount and annual revenue data for BeFit Gyms: one chart that displays the
headcount for each location and one chart that displays the annual revenue for each
location. Make sure to create a meaningful chart title for each chart and create
vertical-axes titles that clearly define the units of measurement. Remove unneces-
sary and redundant information to increase the data-ink ratio. Why might these two
charts be easier for the audience to generate insights from the data than the original
clustered column chart in Problem 16?
b. Recall that the audience is most interested in identifying the incremental revenue a
location should expect from a larger headcount. Create a single column chart using
these data that would compare each location based on both headcount and annual
revenue. (Hint: You can create a new metric for each location that measures the
annual revenue per FTE.) Which location generates the greatest annual revenue per
headcount?
Chapter 4
Purposeful Use of Color
CONTENTS

DATA VISUALIZATION MAKEOVER 4-3 CUSTOM COLOR USING THE HSL COLOR
The Science of Hitting SYSTEM
4-4 COMMON MISTAKES IN THE USE OF COLOR
4-1 COLOR AND PERCEPTION IN DATA VISUALIZATION
Attributes of Color: Hue, Saturation, and Unnecessary Color
Luminance Excessive Color
Color Psychology and Color Symbolism Insufficient Contrast
Perceived Color Inconsistency Across Related Charts
4-2 COLOR SCHEMES AND TYPES OF DATA Neglecting Colorblindness
Categorical Color Schemes Not Considering the Mode of Delivery
Sequential Color Schemes SUMMARY
Diverging Color Schemes
GLOSSARY
PROBLEMS

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Describe and differentiate between hue, saturation, LO 4 Use the hue, saturation, luminance (HSL) system to
and luminance define colors in data visualization software
LO 2 Describe the differences between color psychology LO 5 Employ color to create data visualizations that are
and color symbolism and explain how each can be easier for the audience to interpret
used effectively
LO 6 List common mistakes made when using color in
LO 3 Design color schemes that are appropriate for cat- data visualizations and know how to avoid them
egorical data, ordered data, and quantitative data
with meaningful reference values
 Data Visualization Makeover 129

D ata V is u ali z ation M a k eover

The Science of Hitting

Ted Williams played left field for the Boston Red Sox display; if the batting averages were not included in
from 1939 to 1960, with interruptions for military ser- the circles at the various pitch locations, it would be
vice during World War II and the Korean War. Williams impossible to discern which pitch locations were most
was a 19-time All-Star, a six-time American League favorable for Williams.
batting champion, and Major Leagues Baseball’s last The purpose of the cover art is to show the pro-
.400 hitter. He won the American League Most Valu- gression of a single variable (batting average) from
able Player Award twice. He finished his playing career high to low, so using a single color and creating a
with a .344 batting average and 521 home runs, and gradient effect to define the scale would be more
his lifetime .482 on-base percentage is the highest effective. In our initial revision of the cover art (shown
of all time. Williams is widely regarded as one of the in Figure 4.1b), we replace the various colors used in
greatest hitters in baseball history. the circles of original cover art with red, and we use
In 1970, Williams published the influential book a gradient of red in a manner that corresponds with
The Science of Hitting. In this book, he wrote about the batting averages. The circles corresponding to the
his approach to hitting. The original cover art (shown highest batting averages are relatively dark, and the
in Figure 4.1a) features Williams in his batting stance circles corresponding to the lowest batting averages
with his adjacent strike zone divided into 11 rows and are relatively light. The resulting revised display more
seven columns of nonoverlapping circles representing effectively and efficiently communicates Williams’s
baseballs. Each circle contains Williams’s estimate of performance on pitches in various locations.
his batting average (proportion of at bats in which he Our first revision differs from the original cover art
got a hit) on pitches in that location. By adding colors in a few other subtle but important ways. First, the
to these circles to indicate Williams’s relative batting original cover art features red dashed lines around
average in each pitch location, the book created one the region of pitch locations in which Williams esti-
of the earliest visual displays of a baseball player’s mates he achieves his highest batting averages and
performance. gray dashed lines around the region of pitch locations
Through the original cover art, we can see that in which Williams estimates he achieves his lowest
Williams achieved his highest batting averages on batting averages. The gradient of the color of the
pitches that are around belt-high and across the mid- baseballs in the pitch locations in the first revision
dle of home plate (indicated in red and orange), and communicates the information that these lines are
he achieved his lowest batting averages on pitches intended to convey, and so we have removed the
that are below his belt and on the outside far-right lines. Second, the original cover art features white
portion of home plate (indicated in gray). lines drawn through the baseballs in the pitch loca-
The original cover art communicates the pitch tions on the perimeter of the strike zone. Since the
locations for which Williams achieved his highest and only pitch locations in the display are the pitch loca-
lowest batting averages, and it changed the way many tions in Williams’s strike zone, these white lines are
baseball professionals thought about hitting. However, unnecessary, and we have removed them. Finally, the
with a few small changes, the original cover art could top and bottom halves of circles in four pitch locations
communicate more efficiently and effectively. in the original cover art have different colors. Since we
The original cover art of The Science of Hitting are only given Williams’s estimated batting average
features a single variable—Williams’s estimated bat- for the entire circle in each pitch location (and not the
ting average—for different pitch locations in his strike distinct batting averages in the top and bottom halves
zone. It also uses several different colors to indicate of these locations), this is potentially confusing.
his batting averages in various pitch locations. As the Although our first revision is superior to the orig-
batting average drops, the colors used move from red inal cover art, the values of Williams’s estimated
for the highest batting average to orange to yellow batting averages that are given in each of the pitch
to green to blue to purple and finally to gray for the locations are difficult to read. Further, they convey
lowest batting averages. These colors do not naturally the same information as the gradient of the color, so
lead the viewer to an understanding of the graphical we removed the numerical batting averages in our
(Continued)
130 Chapter 4 Purposeful Use of Color

second revision (Figure 4.1c). Our choice of Figure various locations in the strike zone, the revisions are
4.1b or Figure 4.1c depends on how critical it is for easier to interpret and understand. By observing some
the audience to know the exact value of the estimated basic principles of the use of color and simplifying the
batting averages in the pitch locations. original display, we have created revised displays that
Although the original The Science of Hitting cover communicate the message more effectively and effi-
art was groundbreaking and helped readers under- ciently. In this chapter, we will elaborate on these and
stand how Ted Williams performed on pitches in other guidelines for the effective use of color in charts.

Figure 4.1 Versions of Cover Art for Ted Williams’s The Science of Hitting

Color is one of the Color is the property of an object that results from the way the object reflects or emits
preattentive attributes
light. It is a ubiquitous characteristic, sometimes natural and sometimes by human
discussed in Chapter 3.
design, of virtually every object around us. Color can catch and hold someone’s attention,
communicate, and evoke memories and emotional reactions. This makes color a power-
ful tool that can enhance the meaning and clarity of data visualizations, but to use color
effectively, we must understand how color works and what color can and cannot do well.
In this chapter, we provide the basis for understanding color and using it to create more
effective charts.

4-1 Color and Perception

One of the strengths of color as a means of communication through charts is its ability
to evoke emotions and reactions, create moods, and attract attention to key aspects of the
chart. Understanding how color is perceived enables us to use it to more effectively deliver
our message. Failure to understand how color is perceived can lead to its misuse, which can
result in confusion and unnecessary clutter. In this section, we discuss important consider-
ations in the use of color to effectively communicate through charts.

Attributes of Color: Hue, Saturation, and Luminance

Other color models include One of the three attributes of a color is its hue, which is the base of a color. The primary
Red, Yellow, Blue (RYB) and hues are the three hues that cannot be mixed or formed by any combination of other hues,
Cyan, Magenta, Yellow (CMY).
and all other hues are derived from these three hues. Excel uses the red, green, blue (RGB)
color model with primary hues red, green, and blue. Combinations of these hues create
 4-1 Color and Perception 131

other secondary colors, such as orange, yellow, and violet, and various tertiary colors. The
relationships between primary, secondary, and tertiary colors are commonly displayed on a
color wheel. Figure 4.2 shows the RGB model and RGB color wheel.

FIGURE 4.2 The RGB Primary Color Model and Color Wheel

In addition to hue, colors are commonly distinguished by saturation and luminance.

Saturation refers to the amount of gray in a color and determines the intensity or purity
of the hue in the color. A completely pure hue has no grayness and is referred to as 100%
saturated. As the saturation level decreases, the hue becomes less intense and more grayish.
At 0% saturation, all hues become gray. Figure 4.3 shows the primary hues in the RGB
primary color model at various levels of saturation.

FIGURE 4.3 Primary Hues in the RGB Primary Color Model at Different
Levels of Saturation (at 50% Luminance)

Luminance measures the relative degree of black or white within a color. Adding white
to a hue creates a lighter color, and adding black to a hue creates a darker color. Greater
differences in luminance of a color create greater contrast between them, so luminance is a
good way to indicate hierarchy or degree. However, it is important to note that the human
eye can only discern about 6 or 7 degrees of luminance in a color. At 100% luminance, all
hues become white; at 0% luminance, all hues become black. Figure 4.4 shows the primary
hues in the RGB primary color model at various levels of luminance.
132 Chapter 4 Purposeful Use of Color

FIGURE 4.4 Primary Hues in the RGB Primary Color Model at Different
Levels of Luminance (at Full Saturation)

Combinations of hue, luminance, and saturation determine base, brightness, and gray-
ness of a color. As we will see later in this chapter, Excel allows you to control the hue,
saturation, and luminance of the color of an Excel object.

Color Psychology and Color Symbolism

Color psychology is the study of the innate relationships between color and human behav-
ior. Although human psychological reaction to various colors is not uniform, research
suggests that people react with a high degree of consistency to various colors. For example,
blue skies are thought to make us happier and more energized, while gray skies are thought
to make us sadder and more lethargic.
The psychological categorization of colors as warm and cool is thought to be of particular
importance. Cool hues are considered to be soothing, calming, and reassuring. On the other
hand, warm hues evoke energy, passion, and danger. Purple, blue, and green hues are generally
considered to be cool, and yellow, orange, and red hues are generally considered to be warm.
Color symbolism refers to the cultural meanings and significance associated with color.
Although they are similar and it is sometimes difficult to discern between them, color
psychology and color symbolism are distinct types of stimuli. Color psychology deals
with instinctive relationships between color and human behavior, and color symbolism
refers to learned relationships between color and human behavior. This implies that color
symbolism can differ much more across cultures and can change over time. For example,
blue symbolizes masculinity in Europe and North America, but it symbolizes femininity in
China. People associate green with envy in the United States, but yellow symbolizes envy
to the French and Germans. Yellow is associated with success and power in many African
cultures, and it symbolizes refinement to the Japanese. The relationships between color and
human behavior that are due to color psychology are much more pervasive and reliable than
the relationships between color and human behavior that are due to color symbolism, and
we must be careful when using color symbolism in selecting the color palette for a chart.
Although color psychology and color symbolism are distinct, their ramifications are sim-
ilar. When selecting colors for a visual display, it is important to recognize that color can
enhance or change audience perception, so you should carefully consider the consistence of
the traits associated with the various colors you are considering with your message. Figure 4.5
provides a summary of various traits that have been associated with different colors.1

Perceived Color
We do not perceive color in an absolute manner; that is, we may perceive a color dif-
ferently depending on the other colors currently in our field of vision. Although the

1
Source: nickkolenda.com
4-1 Color and Perception 133

FIGURE 4.5 Traits Frequently Associated with Various Colors

COLOR MEANINGS AND ASSOCIATIONS

Anxiety, Arousing, Daring, Dominant, Energy,

Red Excitement, Health, Life, Love, Passion, Power,
Protection, Spirited, Stimulating, Strength, Up-To-Date

Abundance, Arousing, Comfort, Daring, Excitement,

Orange Extraversion, Fun, Happiness, Lively, Security,
Sensuality, Spirited, Warmth

Arousing, Cheerful, Confidence, Creativity, Excitement,

Yellow Extraversion, Friendliness, Happiness, Optimism,
Self-Esteem, Sincerity, Smiley, Spirited

Calm, Comfort, Equilibrium, Harmony, Health, Hope,

Green Nature, Outdoorsy, Peace, Prosperity, Relaxation,
Security, Serenity, Soothing, Tender

Calm, Comfort, Competence, Coolness, Dignified Duty,

Efficiency, lntelligence, Logic, Peace, Reflection,
Blue
Relaxation, Reliability, Security, Serenity, Soothing,
Successful, Tender, Tranquility, Trust

Authenticity, Charming, Dignified, Exclusive, Luxury,

Purple Quality, Regal, Sensuality, Sophistication, Spiritual,
Stately, Upper Class

Charming, Cheerful, Feminine, Gentle, Nurturing,

Pink
Sincerity, Soft, Sophistication, Tranquility, Warmth

Nature, Outdoorsy, Reliability, Ruggedness, Security,

Brown Support, Tough

Dignified, Efficiency, Elegance, Emotional Safety,

Black Glamour, Power, Richness, Ruggedness, Security,
Sophistication, Stately, Substance, Tough, Upper Class

Calm, Clarity, Cleanness, Down-to-Earth, Happiness,

White Heavens, Honest, Hygience, Innocence, Peace, Purity,
Serenity, Sincerity, Soothing, Tender

wavelengths of light our eyes receive from an object do not change, we perceive the color
of the object in contrast to the colors adjacent to the object. In the rounded rectangle in
the upper left corner of Figure 4.6, the background is a gradient of blue that progressively
darkens from left to right. This rounded rectangle also contains five squares of the same
blue color placed horizontally across the progressive blue gradient. Although the five
squares in this rounded rectangle are the same hue, saturation, and luminance, the squares
appear darker against the lighter portions of the progressive gradient blue background, and
they appear lighter against the darker portions of the progressive gradient blue background.
We perceive the colors this way because of the changes in the contrast between the con-
sistently colored squares and the progressively darkening background. As Figure 4.6 dem-
onstrates, this phenomenon is not a unique characteristic of shades of blue. Using a white
or gray background or light-colored background mitigates this phenomenon and is why a
white or gray background is most commonly used for charts.
134 Chapter 4 Purposeful Use of Color

FIGURE 4.6 Perceived Color Changes as Its Contrast with the

Background Color Changes

Note in Figure 4.6 that as the cool-colored (blue and green) backgrounds become
darker, the squares appear to advance into the foreground of the image. Similarly, as the
warm-colored (red and orange) backgrounds become lighter, the squares appear to recede
into the background of the image. This is because the warmth of a color has an effect on
how close objects appear; warmer colors appear to be nearer and tend to advance into the
foreground. Cooler colors appear more distant and tend to recede into the background.
The same effect can occur when adjacent objects are different colors as well. Saturation
and luminance can also either enhance or diminish the effects of cool and warm colors on
human reactions and behavior.
This effect can be amplified through the use of colors that are directly opposite
each other on the color wheel; such color pairs are known as complementary colors.
Complementary colors create color dissonance; such colors appear stark and create strong
contrast when overlaid or adjacent to each other in a chart. Complementary colors are
useful when you want particular objects in a display to stand out, but overuse of comple-
mentary colors can be distracting for the audience. As shown in Figure 4.2, examples of
complementary colors in the RGB primary color model include blue and orange, green and
red, and yellow and violet.
Colors that are directly adjacent to each other on a color wheel are called analogous
colors. Because of the underlying similarities in their hues, analogous colors appear softer
and smoother than complementary colors when used together. Note that the nearer colors
are to each other on a color wheel, the more analogous they are considered. Analogous
colors appear less stark and create less contrast when overlaid or adjacent to each other
in a data visualization than complementary colors, but their overuse can still be distracting
for the audience. An example of analogous colors in the RGB color wheel shown in
Figure 4.2 is blue and blue-green or blue-violet; another example is orange and red-orange
or yellow-orange.
 4-2 Color Schemes and Types of Data 135

Not all use of color on charts is effective, and sometimes color is pointless, distracting,
or confusing. Unnecessary use of color creates clutter and is potentially detrimental to
charts as it can interfere with the audience’s ability to interpret and understand the mes-
Cognitive load is discussed sage. This is because clutter can increase the audience’s cognitive load, or the amount of
in Chapter 3. effort necessary to accurately and efficiently process the information being communicated
by a chart.
This suggests that when working with color, we must recognize that:
●● Using a multicolored background for data visualization can be distracting.
●● Cool colors appear to be more distant that warm colors.
●● Using complementary colors together creates color dissonance. That is, complemen-

tary colors are said to clash. We use complementary colors to make strong
color-based preattentive distinctions between elements of a data visualization.
Overuse of complementary colors can be distracting for the audience.
●● Using analogous colors creates less color dissonance/more color harmony than

complementary colors. We use analogous colors when we want to make moderate

color-based preattentive distinctions between elements of a data visualization. Over-
use of analogous colors can also be distracting for the audience.
●● Poor use of color can be confusing and distracting to the audience and can increase

the audience’s cognitive load.

Excel provides easy-to-use color palettes that allow you to emphasize contrast using mostly
complimentary colors or monochromatic palettes that provide less contrast by using a
single hue with different levels of luminance and saturation.

4-2 Color Schemes and Types of Data

The color scheme is the set of colors (hues, saturations, and luminances) that are to be used
in a data visualization or a series of related data visualizations. The color scheme that we
select for a visualization should result from strong consideration of the nature of the data we
want to represent with color and the message we want to convey to the audience. For exam-
ple, color can be used in different ways to represent a categorical variable depending on
whether its values represent unordered or ordered groups. When considering a quantitative
variable, the way to use color depends on whether we want to express the magnitudes of the
values or convey how far the values are below or above a predefined reference value (such
as 32° Fahrenheit for temperature). In this section, we consider color strategies for represent-
ing categorical variables with unordered groups, variables with values that can be ordered,
and quantitative variables for which we want to show deviations from a reference value.

Categorical Color Schemes

Categorical variables are Because the values of a categorical variable represent discrete groups, displays of a cate-
discussed in Chapter 1. gorical variable are generally used to communicate information about the absolute or rela-
tive frequency for each group. When the groups of the categorical variable have no inherent
ascending or descending order, the variable is well suited for representation by a distinct
color for each of its unique groups. This type of color scheme is referred to as a categorical
color scheme or a qualitative color scheme.
Because the color assigned to each unique group must appear distinct to the audience,
we generally limit the categorical color scheme to six or fewer colors. When we exceed
six colors, the audience may find distinguishing between groups by the associated color
to be challenging. Figure 4.7 shows examples of categorical color palettes available in
Excel.
136 Chapter 4 Purposeful Use of Color

FIGURE 4.7 Categorical Color Schemes Available in Excel

Colorful
A number of color palettes
useful for categorical data
are available in Excel. To set
the Colorful palette shown
in Figure 4.7, click on the
Page Layout tab, click on the
Colors dropdown menu in the
Themes group, and select the
Office theme.

We can use color to add information on categorical variables to many types of charts.
Consider the data shown in Figure 4.8. The data are zoo attendance by month in two categories,
children and adults. We display these data using a stacked column chart as shown in Figure 4.9.
The following steps maybe used to create this chart.
Step 1. Select cells A1:C13
Click the Insert tab on the Ribbon
Step 2. Click the Insert Column or Bar Chart button in the Charts group
Step 3. When the list of column and bar charts subtypes appears, click the Stacked
Column button
After some editing, the stacked column chart is as appears in Figure 4.9. The first palette
shown in Figure 4.7 was used by default and uses blue and orange, two
complementary colors, to clearly distinguish between the two categories of adult and chil-
dren attendance. To change to another row of the Colorful palette shown in Figure 4.7,
click the chart, click the Chart Design tab and select Change Colors in the Chart
Styles group, and choose the desired palette.
Change
Colors

FIGURE 4.8 Data in the File ZooAttendance

ZooAttendance
 4-2 Color Schemes and Types of Data 137

As we see in Figure 4.9, total zoo attendance is highest in the summer months of
June, July, and August. There is a rather large increase in attendance from November to
December, which is composed of more adults than children. The zoo has its Festival of
Lights in December, which caters to an adult audience.

FIGURE 4.9 A Stacked Column Chart of Children and Adult Zoo Attendance

Zoo Attendance Peaks in Summer

Attendance
25,000
Children Adults

20,000

15,000

10,000

5,000

–
Jan Feb Mar Apr May Jun July Aug Sept Oct Nov Dec

Sequential Color Schemes

When the values of a variable can be arranged in ascending or descending order, a
sequential color scheme (or sequential color palette) should be used. In a sequential
color scheme, the relative degree of saturation or luminance of a single hue is used
to create a gradient that represents the relative value of the variable. Although either
the relative degree of saturation or luminance can be used to create a sequential color
scheme, luminance is most frequently used for this purpose. In Figure 4.10, we show
examples of sequential color palettes available in Excel.

FIGURE 4.10 Sequential (Monochromatic) Color Palettes Available in Excel

Monochromatic

As an example of the use of sequential color, let us consider the average annual temper-
ature in degrees Fahrenheit for each of the 50 states in the United States. These data are in
the file AvgTemp, a portion of which appears in Figure 4.11.
138 Chapter 4 Purposeful Use of Color

FIGURE 4.11 A Portion of the Data in file AvgTemp

AvgTemp

In older versions of Excel, The following steps create the choropleth map of the average temperatures shown in
the Map Charts function may
Figure 4.12.
not exist. You can create a
somewhat similar chart using Step 1. Select cells A1:B51
Excel’s Power Map function.
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Map Chart button in the Charts group and click on
Maps
Filled Map
Step 4. Click Chart Title and enter Average Temperature by State using font
Calibri 16 pt. bold

FIGURE 4.12 Choropleth Map of Average Annual Temperature by State Using Blue

Average Temperature by State

Avg. Temp
70.7

26.6
 4-2 Color Schemes and Types of Data 139

If you have set the Theme to Office, the default of blue will appear (as it is the first row
in Figure 4.10 ). Here, darker blue signifies higher average temperature.
Since blue is a cool color, we might want to switch to a warmer color to help the audience’s
intuition. The following steps allow us to change the color using the Monochromatic col-
ors displayed in Figure 4.10 to produce the map shown in Figure 4.13.
Step 1. Click on the map
Step 2. Click the Chart Design tab on the Ribbon
Step 3. Click the Change Colors drop-down menu in the Chart Styles group
Select the second row in the Monochromatic themes (brown )
as shown in Figure 4.10
In Figure 4.13, we see that higher average temperatures exist in the south where the darkest
color exists. In the lower left, we see that Alaska on the left has a low average temperature
and Hawaii on the right has a high average temperature.

FIGURE 4.13 Choropleth Map of Average Annual Temperature by State Using Brown

Average Temperature by State

Avg. Temp
70.7

26.6

Diverging Color Schemes

When working with a quantitative variable for which there is a meaningful reference value,
such as a target value or the mean, a diverging color scheme (or diverging color palette)
should be used. A diverging color scheme is essentially a gradient formed by the combina-
tion of two sequential color schemes with a shared endpoint at the reference value. These
color schemes use two hues, one of which is associated with values below the reference
value and the other of which is associated with values above the reference value. As the value
of the variable increases, the luminance of the hue associated with values below the refer-
ence value progressively increases and the color becomes lighter until we cross the reference
point. At that point, the luminance of the hue associated with values above the reference
140 Chapter 4 Purposeful Use of Color

point progressively decreases and the color becomes darker. Thus, the hue communicates the
direction of deviation from the reference point, and the luminance conveys the relative devi-
ation from the reference point. For this reason, the hues used on each side of the reference
point in a diverging color scheme are typically distinctive; primary hues are often used to
make it easier to distinguish the direction and degree of deviation from the reference point.
Diverging color schemes are most effective when highlighting both extremes (high
and low values) of a variable. Continuing with another temperature example, consider
the monthly mean daily low Fahrenheit temperatures for Indianapolis for each year from
2010–2019 in Table 4.1.2

TABLE 4.1 Monthly Mean Daily Low Temperatures for Indianapolis 2010–2019
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
2010 18 20 36 49 57 67 70 69 58 46 34 19
2011 17 25 35 45 55 65 72 66 57 45 41 31
2012 26 29 46 44 58 63 72 64 56 44 31 32
2013 22 23 28 42 56 63 65 65 59 45 31 23
2014 11 14 25 42 53 64 61 65 55 45 28 28
2015 15 11 30 44 57 64 66 63 60 47 39 36
2016 21 26 40 43 53 65 68 70 62 51 39 24
2017 28 33 35 49 53 62 67 63 59 49 35 22
2018 17 29 30 37 62 66 68 67 63 47 31 29
2019 20 26 29 44 55 62 69 65 63 46 30 29

We can use the following steps to build a heat map of these data in Excel.
Step 1.Select cells B2:M11
IndLowTemps Step 2.Click the Home tab on the Ribbon and click Conditional Formatting ≠ Conditional
in the Styles group Formatting

Step 3. Click Color Scales and select More Rules

Step 4. When the New Formatting Rule dialog box opens, in the Edit the Rule
Description box
Select 3-Color Scale from the Format Style drop-down menu
Select a blue hue to be associated with values below the reference point
from the Color drop-down menu in the Minimum column
Select Number from the Midpoint drop-down menu for Type, enter 32
in the Value box, and select white to be associated with values near the
reference point from the Color drop-down menu
Select a red hue to be associated with values above the reference point
from the Color drop-down menu in the Maximum column
Check the Preview bar to ensure you have created the diverging color
gradient you desire
Click OK
This creates a heat map with the values in the cells. To format the values so they do not
appear in the cells:
Step 5. With cells B2:M11 still selected, right-click any cell in B2:M11 and select
Entering the semicolon (;)
Format Cells…
in the Type box formats the
values in the highlighted cells
Click the Number tab, then click Custom in the Category box
so they do not appear in the Enter ; in the Type box
cells. Click OK

2
Data from https://www.usclimatedata.com/climate/indianapolis/indiana/united-states
 4-3 Custom Color Using the Hsl Color System 141

These steps produce the heat map with a diverging color scheme shown in Figure 4.14.

FIGURE 4.14 Heat Map of Monthly Mean Daily Low Temperature for Indianapolis 2010–2019

Note that we have considered In the heat map in Figure 4.14, the reference value is 32° Fahrenheit, blue is the hue asso-
color psychology by selecting
ciated with temperatures below the reference value, and red is the hue associated with tem-
red for warmer temperatures
and blue for cooler
peratures above the reference value. As expected, the coldest months (those months shaded in
temperatures when selecting the most intense blue) are January, February, and December, and the warmest months (those
the hues for this heat map. months shaded in the most intense red) are June, July, and August. We also see that March
and November are apparently transition months; in a few years the mean low temperatures
in March or November were below the reference point (shaded blue), and in a few years the
mean low temperatures in March or November were above the reference point (shaded red).

N otes 1 C omments

Diverging color schemes can be difficult to interpret near the also be indistinguishable when the chart is displayed in grays-
reference point in grayscales. The high levels of luminance of cale (as is often the case in printed material).
the two colors on opposite sides of the reference point can

4-3 Custom Color Using the Hsl Color System

In the previous examples, we used Excel’s color palettes to demonstrate the various types of
color schemes. However, it is also possible to customize the colors used in a chart. We can
directly control the hue, saturation, and luminance (HSL) in Excel through the Colors dialog
box, which allows for control of each of these three color characteristics in the following ways.
Hue: The color’s hue expressed as an integer in the range 0 to 255. The primary and second-
ary colors of the RGB primary color mode using fixed values for saturation and luminance are:

Color Hue
Red 0
Yellow 40
Green 80
Cyan 120
Blue 160
Magenta 200

In Figure 4.15, we illustrate changing hues with fixed levels of saturation and luminance in
the Colors dialog box.
142 Chapter 4 Purposeful Use of Color

FIGURE 4.15 Setting the Hue Parameter using the Colors Dialog Box

You can alter the value of Hue: As the value of the Hue parameter increases, the indicator moves horizontally from
by entering a value from 0 to left to right across the color spectrum control in the Colors dialog box to indicate the
255 in the Hue: box, adjusting
the value with the arrow keys
selected hue.
adjacent to the Hue: box, Sat: The color’s saturation expressed as an integer in the range 0 to 255; higher Sat: val-
or using your cursor to drag ues correspond with more intense or pure color, and lower Sat: values produce increasingly
the indicator horizontally gray shades. Setting Sat: to 0 results in a gray tone regardless of the hue and luminance
across the color spectrum settings. In Figure 4.16, we illustrate changing saturation with fixed levels of hue and lumi-
control.
nance in the Colors dialog box.

FIGURE 4.16 Setting the Sat Parameter using the Colors Dialog Box
 4-3 Custom Color Using the Hsl Color System 143

You can alter the value of Sat: As the value of the Sat parameter increases, the indicator moves vertically from the
by entering a value from 0 to
bottom to the top of the color spectrum control in the Colors dialog box to indicate the
255 in the Sat: box, adjusting
the value with the arrow keys
selected saturation, which alters the grayness/increases the purity of the color.
adjacent to the Sat: box, or Lum: The color’s luminosity expressed as an integer in the range 0 to 255. Setting
using your cursor to drag the Lum: to 255 results in white and setting Lum: to 0 results in black. In Figure 4.17,
indicator vertically across we illustrate changing luminance with fixed levels of hue and saturation in the Colors
the color spectrum control.
dialog box.

FIGURE 4.17 Setting the Lum Parameter using the Colors Dialog Box

You can alter the value of Lum: As the value of the Lum: parameter increases, the indicator moves vertically from the
by entering a value from 0 to bottom to the top of the Luminosity slide control in the Colors dialog box to indicate the
255 in the Lum: box, adjusting
the value with the arrow keys
selected luminance, which reduces and increases the lightness of the color.
adjacent to the Lum: box, or As an example, let us consider again the zoo attendance stacked column chart shown
using your cursor to drag the in Figure 4.18. Suppose we wish to use a different shade of orange to represent the adult
indicator vertically across attendance. The following steps illustrate how we can change the adults category to a cus-
the luminosity slide control. tomized color.
Step 1. Open the file ZooChart
Step 2. Click on the orange portion of any column and then right click
Step 3. Click the Shape Fill button
ZooChart Fill
Select More Fill Colors…
When the Colors dialog box appears, click the Custom tab
Step 4. Next to Color model: choose HSL from the drop-down menu
Step 5. Set Hue to 21, Sat to 238, and Lum to 182
Step 6. Click OK
The zoo data with the lighter shade of orange is shown in Figure 4.18.
In some instances, you may want to replicate a color used in an existing image. You can
use the Eyedropper, a tool in PowerPoint that determines the HSL settings for colors used
in an image. For example, consider an analyst who is creating a presentation to give to the
management of Grappenhall Publishers and wants to use the color scheme of the compa-
ny’s logo (shown in Figure 4.19) in creating this presentation.
144 Chapter 4 Purposeful Use of Color

FIGURE 4.18 A Stacked Column Chart of Children and Adult Zoo Attendance with Lighter Orange

Zoo Attendance Peaks in Summer

Attendance
25,000
Children Adults

20,000

15,000

10,000

5,000

–
Jan Feb Mar Apr May Jun July Aug Sept Oct Nov Dec

FIGURE 4.19 Grappenhall Publishers Logo

The following steps explain how to use the Eyedropper tool in PowerPoint to determine
the values of hue, saturation, and luminance that create the distinctive purple color of the
Grappenhall Publishers logo in Figure 4.19.
Step 1. Open a new PowerPoint document presentation
Step 2. Paste a copy of the image of interest into a blank slide
Step 3. Click the Insert tab on the Ribbon
Click the Shapes button in the Illustrations group, select any shape icon,
and use the cursor to draw the shape into the same slide
Step 4. Right click the shape you have drawn and select Format Shape to open the
Format Shape task pane
On the Format Shape task pane
Click the Fill & Line icon
Under Fill, select Solid Fill
4-3 Custom Color Using the Hsl Color System 145

Click the Color tool under Fill, then select Eyedropper to open
the Eyedropper tool
Drag the Eyedropper tool over the object that has the color you want
to replicate (in this case, the image of the Grappenhall Publishers
logo) until the Eyedropper tool fills with the desired color, then click
The fill color of the shape you drew on the PowerPoint slide will now match the color of
the Grappenhall Publishers logo. The next step determines the values of hue, saturation,
and luminance that can be used to duplicate this color.
Step 5. Click the object that you used the Eyedropper to color
Click the Color tool under Fill, then select More Colors… to open the
Colors dialog box
Click the Custom tab in the Colors dialog box, select HSL for Color
model (Figure 4.20)

FIGURE 4.20 Colors Dialog Box Showing the Values of Hue, Saturation,
and Luminance used in the Grappenhall Publishers Logo

We see in Figure 4.20 that the purple color Grappenhall Publishers uses in its logo is
created with Hue = 194, Sat = 174, and Lum = 63. We can now replicate this specific
color in the data visualizations we create for our presentation.
We can use the same process to identify the values of Hue, Sat, and Lum (211, 192, and
154, respectively) that replicate the pink color of the heart in the Grappenhall logo.
146 Chapter 4 Purposeful Use of Color

N o t e s 1 C o m m e nt s

1. In addition to the HSL color scheme, Excel has the Red, on the New-Current box in the lower right of the dialog box.
Green, Blue (RGB) color scheme available corresponding Clicking OK will change the object to the new color.
to the Red, Green, Blue color wheel. You may access this 3. In Excel’s RGB color model, the Transparency slider in the
by selecting RGB from the Color model: dropdown menu Colors dialog box controls how much you can see through a
in the Colors dialog box. color. You can drag the Transparency slider or enter a number
2. The Standard tab in the Colors dialog box allows the user to between 0 and 100 in the Transparency slider box adjacent to
modify color in Excel's RGB color model. Clicking on a par- the slider. You can vary the percentage of transparency from 0
ticular hexagon invokes the associated color that appears (fully opaque, the default setting) to 100% (fully transparent).

4-4 Common Mistakes in the Use of Color

in Data Visualization
Although color is a powerful tool for communicating with charts, it is often misused. When
misused, it may distort the intended message or distract the audience. In this section, we
discuss several common mistakes committed when using color in a data visualization.

Unnecessary Color
Data visualization experts agree that color should only be used when it communicates
something that no other aspect of a chart communicates to the audience. Figure 4.21 shows
the number of units sold (in thousands) for seven top-selling midsize sedans.
In this chart, the audience can discern which column corresponds to each of the models
through the colors of the columns and the legend. Although this communicates the data,
 4-4 Common Mistakes in the Use of Color in Data Visualization 147

we can accomplish the same communication with a chart that creates less cognitive load by
avoiding the use of multiple colors. If we clearly label the columns on the horizontal axis,
then there is no need for a different color for each model of sedan.

FIGURE 4.21 Use of Hue and a Legend for Identification of Categories

US Unit Sales for Top Selling Midsize Sedans

Sales (1000s)
30
Chevrolet Malibu
25
Ford Fusion
20 Honda Accord

15 Hyundai Sonata
SedanSalesChart
Kia Optima
10 Nissan Altima
5 Toyota Camry

The chart shown in Figure 4.22 uses only horizontal-axis labels to communicate the
association between the columns and the models. The audience now does not have to look
back and forth between the columns and the legend to make these associations in this chart,
which reduces the audience’s cognitive load.

FIGURE 4.22 Use of Horizontal Axis Labels for Identification of Categories

US Unit Sales for Top Selling Midsize Sedans

Sales (1000s)
30

25

20

15

10

0
Chevrolet Ford Fusion Honda Hyundai Kia Optima Nissan Toyota
Malibu Accord Sonata Altima Camry
Model

Note that one could produce a chart that includes both the colors and legend of Figure 4.21
and the horizontal-axis labels of Figure 4.22, but this would embed redundant information in
the chart and further decrease its data-ink ratio.
148 Chapter 4 Purposeful Use of Color

Excessive Color
There is a limit to the amount of information that can be communicated to the audience
using color. Suppose you are analyzing quarterly house-price indexes from 1992–2019 for
each state and the District of Colombia, and you want to emphasize the westernmost states
in the continental United States (Arizona, California, Nevada, Oregon, and Washington).
The chart in Figure 4.23 shows the quarterly house-price index from 1992–2019 for each
state and the District of Colombia.3 The chart captures information on house-price index
by state for each quarter of the 28-year period (112 quarters). It enables the audience to
quickly see that on a national level house prices increased steadily until around 2004, when
the rate of increase accelerated dramatically until sometime around 2007. The audience can
also see that the housing market then crashed and house prices generally fell for about four
StateHousingIndicesChart to five years, until they began to increase again sometime around 2010.

FIGURE 4.23 An Example of Too Many Colors on One Chart

Quarterly House Price Index by State (Base = 1991) AK AL

Home Price Index AR AZ
CA CO
600 CT DC
DE FL
GA HI
IA ID
500 IL IN
KS KY
LA MA
MD ME
400
MI MN
MO MS
MT NC
300 ND NE
NH NJ
NM NV
NY OH
200 OK OR
PA RI
SC SD
TN TX
100 UT VA
VT WA
WI WV
0 WY
1999 2004 2009 2014 2019
Year

You can re-create the chart in the file StateHousingIndicesChart using the file
StateHousingIndices and the following steps:
Step 1. Open the file StateHousingIndices and select the data for AK in cells A2:D85
Use the Recommended Charts button in the Charts group on the
Recommended
Charts

Insert tab of the Ribbon to create a line chart using these data
StateHousingIndices
Step 2. Right-click the chart, then select Select Data to open the Select Data Source
dialog box (shown in Figure 4.24)
Click Series1 in the Legend Entries (Series) area and click the Remove
button Remove to remove this series from the chart
Click Series2 in the Legend Entries (Series) area
Click the Edit button Edit in the Legend Entries (Series) area to

open the Edit Series dialog box for this series

3
Data from https://www.fhfa.gov/DataTools/Downloads/Pages/House-Price-Index-Datasets.aspx
4-4 Common Mistakes in the Use of Color in Data Visualization 149

FIGURE 4.24 Select Data Source Dialog Box

Click the Series name: box, then select cell A2 to use the contents of
this cell (AK) for the series name
Click OK to close the Edit Series dialog box
Click OK to close the Select Data Source dialog box
Step 2. Right-click the legend at the bottom of the chart, then click Format Legend…
to open the Format Legend pane
Click Legend Options, then click the Legend Options button and
select Right
This produces the chart in Figure 4.25.

FIGURE 4.25 Line Chart of Index Values for AK

AK
300.00

250.00

200.00

150.00
AK
100.00

50.00

0.00
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019

To add the next state in the data (AL) to the chart:

Step 3. Right-click the chart and select Select Data… to open the Select Data Source
dialog box
150 Chapter 4 Purposeful Use of Color

Click the Add button Add in the Legend Entries (Series) area to open

the Edit Series dialog box for this new series

Click the Series name box and select cell A86 to use the contents of
this cell (AL) for the series name
Click the Series values box and select cells D86:D169 to use the
contents of these cells for this new series (AL)
Click OK
Click the Edit button Edit in the Horizontal (Category) Axis Labels
area to open the Axis Labels dialog box
Click the Axis label range box and select cells B86:B169
Click OK
Click OK
This produces the chart in Figure 4.26.

FIGURE 4.26 Line Chart of Index Values for AK and AL

300.00

250.00

200.00

150.00 AK

AL
100.00

50.00

0.00
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019

To add the remaining states and the District of Columbia to the chart:
Step 4. Repeat Step 3 for each remaining state and the District of Columbia.
This and some editing produces the chart in Figure 4.23.
Comparing housing prices across states with this chart is difficult because there are
51 categories/colors on this chart. The human brain can only process a few distinct colors simul-
taneously. The use of a legend in Figure 4.23 also creates additional eye travel for the audience.
We can make it easier for the audience to find the westernmost states in the continental
United States (Arizona, California, Nevada, Oregon, and Washington) on this chart by
applying a different color for each of these states and using gray for all other states. We can
also remove the legend and add a label for each of the five westernmost states in the conti-
nental United States to the end of their respective lines. The following steps detail how to
produce the chart that results from these modifications.
Step 1. Open the file StateHousingIndicesChart (or use the chart that you created with
StateHousingIndicesChart the file StateHousingIndices)
Step 2. Double click AZ in the legend of the chart to select this series and open the
Format Legend Entry task pane (make sure you have selected only “AZ” and
not selected the entire legend)
 4-4 Common Mistakes in the Use of Color in Data Visualization 151

Step 3. When the Format Legend Entry task pane opens

Click the Fill & Line icon
Select Solid Line in the Border area
From the drop-down Color menu, click More Colors…
In the Colors dialog box, click on the Custom tab
From the Color model: drop-down menu, select HSL
Change the value of color of the line associated with AZ (Figure 4.27 uses
Hue = 32, Sat = 255, Lum = 128)
Step 4. Repeat Steps 2–3 for
CA (Figure 4.27 uses Hue = 155, Sat = 133, Lum = 132)
NV (Figure 4.27 uses Hue = 0, Sat = 255, Lum = 128)
OR (Figure 4.27 uses Hue = 112, Sat = 173, Lum = 71)
WA (Figure 4.27 uses Hue = 222, Sat = 230, Lum = 92)
Step 5. Repeat Steps 2–3 for each other state and DC using the same shade of gray
(the chart in Figure 4.24 uses Hue = 0, Sat = 5, Lum = 207)
Step 6. Right-click AZ in the legend of the chart to select this series and select Format
Data Series…
When the Format Data Series task pane opens, click the Series Options
icon
Select Secondary Axis to put the AZ series on a secondary vertical axis that
will be located on the right side of the chart and may overlap the legend
Step 7. Repeat Step 7 for each of the CA, NV, OR, and WA series
Step 8. Click the legend and press Delete
Step 9. Click the secondary legend to open the Format Axis task pane
Click the Axis Options icon
Enter 600 for Maximum under Bounds in the Axis Options area so the
scales are the same on the primary and secondary vertical axes
Step 10. Click the secondary vertical axis and press Delete
Using a secondary axis in Step 11. Click any area in the chart
Steps 7–10 brings the lines
Click the Insert tab on the Ribbon
associated with the states in A

which we have an interest to

Click the Text box button in the Text group, and use the cursor to draw
Text
Box

the forefront of the chart. the shape into the same slide adjacent to the AZ line, and enter Arizona
into the text box
Click the text box and format its contents (the chart in Figure 4.24 uses
Calibri 9pt font with Hue = 32, Sat = 255, Lum = 128)
Step 12. Repeat Step 11 for the CA, NV, OR, and WA series (entering California, Nevada,
Oregon, and Washington in the respective textboxes). Use Calibri 10.5pt font
with the values of hue, sat, and lum for each of these states that is given in Step 4
It is much easier to see in Figure 4.27 that the five westernmost states experienced more rapid
growths in housing prices than did most other states immediately prior to the 2007 housing
market collapse, and the drops in housing prices in these states during the 2007 housing market
collapse were steeper than most other states. The chart also shows that the states in this region
have experienced an average, or stronger than average, post-housing market collapse recovery.

Insufficient Contrast
When using color to distinguish between elements of a chart, it is important that the audi-
ence can easily distinguish between the selected colors. If the colors assigned to different
chart elements are difficult to differentiate, the audience can become confused or may have
to work harder than necessary to understand the chart’s message.
In a story on the change in the portion of England’s public services budget devoted to the
country’s National Health Service (NHS) over a 60-year period, the British Broadcasting
Company (BBC) included the pie charts originally produced by the Institute for Fiscal
Studies as shown in Figure 4.28. The charts use color to differentiate between spending on
the NHS (hue = 127, sat = 241, lum = 40) and the remainder of the public services budget
(hue = 127, sat = 241, lum = 53). The use of green hues takes advantage of the color’s
152 Chapter 4 Purposeful Use of Color

FIGURE 4.27 Quarterly House Price Index by State with More Efficient Use of Color

Quarterly House Price Index (Base = 1991) for Western States Demonstrates More Volatility
Home Price Index
600

500

Oregon

400 Washington
Arizona

300 California
Nevada

200

100

0
1999 2004 2009 2014 2019
Year

common symbolic association with finance and economics, but these colors only differ
slightly in luminance. This difference is not sufficient to allow the audience to easily dis-
cern which slice is associated with spending on the National Health Service and which slice
is associated with the remainder of the public services budget in each of these pie charts.

FIGURE 4.28 Lack of Contrast Makes It More Difficult to Differentiate

between Slices in the Pie Charts

As discussed in Chapters 2
and 3, the use of pie charts is
generally discouraged.
 4-4 Common Mistakes in the Use of Color in Data Visualization 153

A simple adjustment to the color used to represent the remainder of the budget will
make these pie charts far easier for the audience to process with less cognitive load. In
Figure 4.29, we show charts re-created in Excel using a hue level of 127 and saturation
level of 241 for both categories, and a luminance level of 131 for the NHS and 53 for the
rest of the budget to increase the visual contrast between the categories.

FIGURE 4.29 Increased Contrast Makes It Easier to Differentiate between

Slices in the Pie Charts

The most extreme and most Because we have increased the contrast between the two slices of each pie chart by
confusing form of insufficient
increasing the difference in the luminance, the audience can more easily differentiate
contrast is redundancy, which
occurs when the same color is
between the NHS portion of the budget and the rest of the budget for each time period in
associated with two unrelated the display. The use of a higher luminance for the NHS portion of the budget also focuses
categories or variables in a the attention on that portion of the budget.
chart or series of charts.

Inconsistency across Related Charts

Data dashboards are When creating several charts for a single report, a presentation, an ongoing analysis, or a
discussed in Chapter 8.
data dashboard, it is critical that color is used consistently. Using different color schemes
or using different colors to represent categories on different charts will confuse the audi-
ence and dramatically increase its cognitive load.
Consider the zoo attendance data shown in Figure 4.8, shown again here in Figure 4.30.
The column chart uses orange and blue, two complementary colors, to distinguish between
adult and children. Now suppose as part of a presentation to the Zoo Board, you have
Figure 4.30 on a slide, and it will be followed by a second slide that shows adult versus
children December ticket revenue for the last five years. Consistency in the use of color
across slides will make it easier for the audience to comprehend. So, rather than use differ-
ent colors, it is helpful to use the same orange and blue to represent other factors related
to adults and children. Figure 4.31 shows a chart of the December revenue data. The con-
sistency in the use of color will help the audience’s comprehension of the data presented
across the series of slides.

Neglecting Colorblindness
Consider the monthly mean daily low temperatures for Indianapolis 2010–2019 in Table
4.1. If we had used green to represent temperatures below freezing in the gradient we
developed for Figure 4.14, the result would look like the chart in Figure 4.32.
154 Chapter 4 Purposeful Use of Color

FIGURE 4.30 A Stacked Column Chart of Children and Adult Zoo Attendance

Zoo Attendance by Month

Attendance
25,000
Children Adults

20,000

15,000

10,000

5,000

–
Jan Feb Mar Apr May Jun July Aug Sept Oct Nov Dec

FIGURE 4.31 A Line Chart of Ticket Revenue for the Month of December

Ticket Revenue for the Month of December

Revenue ($) Children Adults

30,000

25,000

20,000

15,000

10,000

5,000

–
2016 2017 2018 2019 2020
Year

Although green may not be as appealing in this chart because it is not as closely asso-
ciated with cold as is blue, the audience can still interpret the heat map in Figure 4.29 with
relative ease. That is, unless a member of the audience is red-green colorblind, in which
case the resulting heat map may look like the chart in Figure 4.33.
 4-4 Common Mistakes in the Use of Color in Data Visualization 155

FIGURE 4.32 Heat Map of Monthly Mean Daily Low Temperature for Indianapolis 2010–2019
Using Green and Red for Temperatures Below and Above Freezing

FIGURE 4.33 Heat Map in Figure 4.32 as It May Appear to Someone Who Is Red-Green
Colorblind

Some of the blocks representing the cold months of December, January, and February in
Figure 4.32 appear to be identical to some of the blocks representing the hot months June,
July, and August for someone who is red-green colorblind, so this chart has little meaning
for those members of the audience who have difficulty differentiating red from green.
Colorblindness, or a reduced ability to accurately perceive some colors, occurs when at
least one of the three types of cones in a retina is insensitive to the wavelength of light it is
responsible for sensing. The most common form of colorblindness is red-green colorblind-
ness, which affects approximately 8% of all men and approximately 0.5% of all women.
People with red-green colorblindness cannot perceive, to some degree, differences between
red and green. Blue-yellow colorblindness is far less common, occurring in approximately
156 Chapter 4 Purposeful Use of Color

.01% of the population. These individuals cannot perceive, to some degree, differences
between blue and yellow. The 0.003% of the population that is completely colorblind see
only in shades of gray. If you neglect to consider color-blindness when you select colors
for your charts, you risk making it more difficult for some members of the audience to
understand your message.

Not Considering the Mode of Delivery

Color works differently in print and projection. Projected presentations will be seen from
a distance, and the audience will generally have relatively little time to review specific
aspects of a projected presentation. Use of thick lines, color contrast, and relatively high
saturation and luminance in this medium is critical. Use sharp outlines and saturated con-
trasting colors when creating a presentation that is to be projected. On the other hand, the
audience will generally be close to printed presentations and will have more time to review
specific aspects of a printed report, so you do not need to rely as extensively on thick lines,
color contrast, and relatively high saturation and luminance in this medium. Use softer out-
lines with colors with less saturation, lower luminance, and less contrast when creating a
presentation that is to be printed.
Colors will also vary from monitor to monitor, projector to projector, and monitor to
projector. When possible, go through your entire presentation in advance with the projec-
tion equipment that will you use when giving the actual presentation, and make necessary
adjustments iteratively until you are satisfied with the projected colors. If this is not possi-
ble, avoid extremely bright colors in your color scheme as those often differ most between
the appearance on the monitor you use when creating the presentation and the appearance
on the screen with the projection system you will ultimately use.
Similarly, colors vary from printer to printer and from monitor to printer. Consider
creating a test page that includes each of the colors you use in your report on the
printer you will ultimately use to print your report. Print this single page, review the
results, adjust as necessary, and repeat the process until you are satisfied with the
printed colors.

N otes 1 C omments

1. It is a good idea to view your data visualizations in grays- 2. Several colorblind-friendly color schemes have been
cale to ensure that the levels of saturation and luminance developed. Popular examples include those developed
used provide sufficient basis for colorblind members of the by the IBM Design Library (https://www.ibm.com/design/v1
audience to understand the intended message. This can be /language/resources/color-library/) and David Nichols
done in Excel by selecting Grayscale setting in the Print (https://davidmathlogic.com/colorblind/).
menu and viewing the result in the Print Preview window.

S U M M A RY

In this chapter, we have discussed specific aspects of the preattentive attribute color and
how to use color to create effective data visualizations. We defined the roles of hue, sat-
uration, and luminance in defining a color. We discussed the differences between color
psychology and color symbolism and how each can be used effectively. We then described
how to design color schemes that are appropriate for categorical variables, ordered vari-
ables, and quantitative variables with meaningful reference values. We discussed the HSL
system for defining colors in data visualization software, and we demonstrated how to use
the HSL system in Excel. We concluded with a discussion of various common mistakes
Problems 157

made when using color in data visualizations, including neglecting to consider colorblind-
ness when creating data visualizations. And throughout the chapter, we provided examples
and step-by-step instructions on how to address the issues discussed in this chapter across a
wide variety of chart types.

G L O S S A RY

Analogous colors Colors that are directly adjacent to each other on a color wheel.
Categorical color scheme A set of colors used to describe a categorical variable when
the categories have no inherent ascending or descending order. Also called a categorical
color palette.
Color symbolism The cultural meanings and significance associated with color.
Cognitive load The amount of effort necessary to accurately and efficiently process the
information being communicated by a chart.
Color A preattentive attribute for data visualizations that includes the attributes hue,
saturation, and luminance.
Color psychology The study of the innate relationships between color and human
behavior.
Color scheme A set of colors used in a chart or in a series of related charts. Also called a
color palette.
Color wheel A common chart used to show the relationships between primary, secondary,
and tertiary hues for a primary color model.
Colorblindness An inability to accurately detect some colors.
Complementary colors Colors that are directly opposite each other on a color wheel.
Cool hues Hues that are thought to be soothing, calming, and reassuring. Blues, purples,
and greens are generally considered to be cool hues.
Diverging color scheme A set of colors used in a chart or in a series of related charts to
describe values of a quantitative variable for which there is a meaningful reference value,
such as a target value or the mean. Also called a diverging color palette.
Hue The attribute of a color that is determined by the position the light occupies on the
visible light spectrum and defines the base of the color.
Luminance The attribute of a color that represents the relative degree of black or white in
the color.
Primary hues The three hues in a primary color model that cannot be mixed or formed by
any combination of other hues. All other hues in a primary color model are derived from
these three hues.
Saturation The attribute of a color that represents the amount of gray in the color and
determines the intensity or purity of the hue in the color.
Sequential color scheme A set of colors used to describe the values of a quantitative
variable or a categorical variable when the categories have an inherent ascending or
descending order. Also called a sequential color palette.
Warm hues Hues that are thought to evoke energy, passion, and danger. Yellows, oranges,
and reds are generally considered to be warm hues.

P ro b l e ms

CONCEPTUAL
1. Understanding Hue, Saturation, and Luminance. For each of the following state-
ments, indicate whether hue, saturation, or luminance is being described. LO 1
a. The amount of gray in a color
b. The position the light occupies on the visible light spectrum
c. The intensity or purity of a color
d. The relative degree of black or white in the color
158 Chapter 4 Purposeful Use of Color

2. Differentiating Between Hue, Saturation, and Luminance. The file HueSatLumChart

contains six worksheets named a, b, c, d, e, and f that each contain a pair of colored
rectangles. The colors of the rectangles in each of these pairs differ on a single charac-
HueSatLumChart
teristic of color (hue, saturation, or luminance). Use this file and the Colors dialog box
to determine whether each pair of colors differs on hue, saturation, or luminance, and
explain your answer. LO 1
a.

b.

c.

d.

e.

f.

3. Using Hue, Saturation, and Luminance to Differentiate Categories. The charts

in parts a, b, and c each show 2019 and 2020 quarter 1 sales for nine popular midsize
sedans. These charts are included in worksheets of the same names (a, b, and c) in the
Q1SalesChart
file Q1SalesChart. Use the Colors dialog box to determine whether hue, saturation,
or luminance is used to differentiate the Honda Accord from the other eight models in
each chart. LO 1
a.
Midsize Sedans Quarter 1 Unit Sales
Sales ($1000s)
90
80
70
60
50
40
30
20
10
0
2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020

Toyota Camry Nissan Altima Honda Accord Ford Fusion Chevrolet Malibu Subaru Outback Tesla Model 3 Kia Optima Hyundai Sonata
 Problems 159

b.
Midsize Sedans Quarter 1 Unit Sales
Sales ($1000s)
90
80
70
60
50
40
30
20
10
0
2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020

Toyota Camry Nissan Altima Honda Accord Ford Fusion Chevrolet Malibu Subaru Outback Tesla Model 3 Kia Optima Hyundai Sonata

c.
Midsize Sedans Quarter 1 Unit Sales
Sales ($1000s)
90
80
70
60
50
40
30
20
10
0
2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020 2019 2020

Toyota Camry Nissan Altima Honda Accord Ford Fusion Chevrolet Malibu Subaru Outback Tesla Model 3 Kia Optima Hyundai Sonata

d. Which of the charts in parts a, b, and c is least effective in differentiating between

the Honda Accord and the other models? Why?
4. Using Categorical Color Schemes, Sequential Color Schemes, and Diverging
Color Schemes. Identify whether a categorical color scheme, a sequential color
scheme, or a diverging color scheme should be applied in each of the following exam-
ples. In each instance, explain why this is the appropriate color scheme. LO 2
a. You have been given the following data on a popular regional product’s unit sales
and net profit for several markets.

City Sales (Units) Net Profit

Amarillo 99,299 $122,619
Ft. Worth 112,219 −$132,762
El Paso 117,893 $169,500
Dallas 139,486 −$136,332
Houston 143,721 $63,646
San Antonio 157,241 $115,632
Austin 175,555 $80,707
160 Chapter 4 Purposeful Use of Color

From these data, you have created the following bar chart to show the number
of units sold by market. Now you want to add the information on net profit to the
bars on this chart to show the profit or loss earned in each market.
Market

Austin

San Antonio

Houston

Dallas

El Paso

Ft. Worth

Amarillo
0 50000 100000 150000 200000
Sales (units)

b. You have been given the following data on mean daily high temperature for a city
and number of cups of hot chocolate a chain of coffee shops sold in that city by
month.

Mean Daily High

Month Temperature in Fahrenheit Unit Sales
Jan 27 144,908
Feb 32 120,032
Mar 43 103,917
Apr 50 90,866
May 62 60,110
Jun 74 36,357
Jul 83 16,002
Aug 77 29,445
Sep 62 63,523
Oct 53 81,847
Nov 40 112,818
Dec 33 130,423

From these data, you have created the following column chart to show the mean
daily high temperature by month for this city. Now you want to add the information
to this chart to show the number of cups of hot chocolate that were sold each month.
Mean Daily High Temperature by Month
Temperature (degrees Fahrenheit)
100

80

60

40

20

0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
 Problems 161

c. You want to add information on the continent of the nation in which headquarters
for the top 50 companies in an industry are located to the following column chart
that shows the number of these headquarters by nation.
Number of Top 50 Firm Headquarters by Nation
Number of Firms
14
12
10
8
6
4
2
0
US France Great Britain Japan Germany China Canada Chile India Brazil South Africa
Nation

d. You have been asked by a real estate agent to use the following data to create a
chart that provides insight into the nature of the relationship between the sales price
and number of days on the market for homes she has recently sold.

Sales Price Days on the Market Neighborhood

$247,412 58 East Highland
$249,787 60 Sebastian
$194,955 34 Hargrove Hills
$173,612 55 Hargrove Hills
$171,299 55 Sebastian
$281,197 62 Sebastian
$264,477 25 Hargrove Hills
$264,838 50 East Highland
$261,819 30 East Highland
$293,953 42 East Highland
$296,002 32 East Highland
$167,136 38 Hargrove Hills
$334,063 33 Hargrove Hills
$332,340 32 Manafort Falls
$264,066 47 Manafort Falls
$255,070 55 Manafort Falls
$185,572 29 East Highland
$314,198 31 East Highland
$256,183 42 Manafort Falls
$230,741 63 Sebastian
$274,552 38 Manafort Falls
$262,332 45 Manafort Falls
$219,969 39 Hargrove Hills
$253,708 41 East Highland
$245,114 55 Sebastian
$296,920 49 Manafort Falls
$295,866 25 Hargrove Hills
$270,353 59 Manafort Falls
$202,422 55 Sebastian
$214,206 53 East Highland
162 Chapter 4 Purposeful Use of Color

You have created the following scatter chart. The real estate agent has asked
you to add the information on the neighborhood in which each of these homes is
located to the scatter chart you have created.
Relationship Between Days on the Market and
Sales Price for Homes
Days on Market
70

60

50

40

30

20

10

0
$0 $100,000 $200,000 $300,000 $400,000
Sales Price

e. You have been given the following data on monthly change in number of subscrib-
ers to several online publications devoted to popular hobbies. You have now been
asked to add color to this display to indicate whether the number of subscribers is
increasing or decreasing each month.

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Sporting
Goods 0.9% −0.7% −0.1% 0.2% 0.2% 0.5% −0.3% 1.6% 1.2% 0.3% −0.8% 0.9%

Philatelist
Quarterly 0.5% 0.1% 0.3% −0.1% 0.5% 0.5% −0.9% −0.1% 0.1% −0.4% −1.2% 0.3%

American
Spelunker 0.2% −1.6% 0.6% 1.6% −1.4% 0.0% −0.1% 0.6% −1.3% −1.2% −0.8% 0.4%

Home Cooks 0.5% −0.9% 0.2% 1.2% 0.8% −0.3% 1.2% 0.5% 0.6% −1.4% −0.6% 0.3%

Classic
Literature
Digest 0.5% 0.2% −0.1% 0.4% 0.0% 0.4% 1.2% 0.0% −0.7% −1.4% −0.1% 1.4%

Numismatics
Magazine −0.1% 0.0% 0.5% 0.4% 0.0% 1.0% −0.7% 1.4% 0.1% 1.2% 1.3% 1.4%

Gardening
World 0.7% −0.5% 0.5% −1.8% 1.5% −0.6% 0.8% −1.2% 0.9% 1.6% 0.9% 1.0%

Knitting News 1.1% 1.2% −1.0% −1.2% −0.3% 1.5% −0.3% 0.6% −0.2% −0.5% 1.2% 1.1%

Online Gamer 1.1% 1.2% 0.7% 0.5% 1.8% 1.1% 0.2% 0.1% 0.4% 1.0% 1.4% 2.1%

Popular
Photographer −0.6% 0.3% −0.9% 0.1% 0.8% 1.4% −0.1% −1.0% −0.4% −0.7% 0.7% 0.8%
Problems 163

5. Considering Colorblindness in the Design of a Chart. Four design artists

(Frances Bittinger, Dorian Martin, Bill Zander, and Mariana Knoop) of a regional
marketing research firm’s graphic arts department have each designed a chart for
use in a presentation to a team of Happy Lawn landscaping service executives,
two of whom are colorblind. Each has designed a heat map showing the relative
percentage change in the number of customers various lawn care services experi-
enced over the past year in several Midwestern states: Illinois (IL), Indiana (IN),
Michigan (MI), Minnesota (MN), Ohio (OH), and Wisconsin (WI).
Explain why each of these charts may or may not be difficult for the colorblind
members of the team of Happy Lawn landscaping service executives to visually pro-
cess. LO 6
Percentage Change in Customers (by Frances Bittinger)

Percentage Change in Customers (by Dorian Martin)

Percentage Change in Customers (by Bill Zander)

Percentage Change in Customers (by Mariana Knoop)

164 Chapter 4 Purposeful Use of Color

6. Identifying Mistakes in the Use of Color in a Column Chart. According to USA

Today, the headquarters of over one-third of all Fortune 500 (F500) companies are
located in one of six U.S. metropolitan areas: Chicago, Dallas-Ft. Worth, Houston,
Minneapolis-St. Paul, New York, and San Francisco. The following chart displays the
number of F500 headquarters for each of these metropolitan areas. LO 6

Explain what color is being used to uniquely communicate in this chart, what
aspects of the use of color in this chart are ineffective, and how the weakness(es) in the
use of color in this chart could be corrected. Also indicate all other revisions you would
make to improve this chart.
7. Identifying Mistakes in the Use of Color in a Bar Chart. The most widely grown lawn
turf grasses in the southeastern United States are warm-season grasses: bahiagrass,
bermudagrass, centipedegrass, St Augustinegrass, and zoysia. Dana Tanner, crew chief
with Holiday Grass lawn care service in Birmingham, Alabama, is now planning her
resource needs for the upcoming summer. Each of these grasses has different fertilizer
and direct sunlight needs, drought and heat tolerance, growth rate, and disease and
weed resistance. Therefore, it is important that Ms. Tanner understand the number of
lawns of each grass in her district when planning her resource needs for the upcoming
summer. To develop a better understanding of the number of lawns of each grass in her
district, she observes and records the type of lawn for a random sample of 500 homes
in her district, and she summarizes her results in the following chart. LO 6
Grasses Grown on Birmingham Area Lawns
Type of Grass

Zoysia

Bermudagrass

St.Augustinegrass

Bahiagrass

Centipedegrass

0 50 100 150 200

Number of Homes
 Problems 165

Explain what information color is being used to uniquely communicate in this chart,
what aspects of the use of color in this chart are ineffective, and how could the weak-
ness(es) in the use of color in this chart be corrected.
APPLICATION
8. Comparing the Number of Fortune 500 Headquarters in Six U.S. Metropolitan
Areas. Consider the number of Fortune 500 (F500) headquarters in the six U.S. metro-
politan areas (Chicago, Dallas-Ft. Worth, Houston, Minneapolis-St. Paul, New York, and
F500MetrosChart
San Francisco) that are home to the greatest numbers of F500 headquarters. Modify the
following column chart (from Problem 6) displaying these data, which is provided in the
file F500MetrosChart, so that it uses color more effectively. Also make all other neces-
sary revisions to improve this chart. Explain why the changes you have made produce a
chart that is superior to the original bar chart from Problem 6. LO 5

9. Frequency of Types of Grass. Consider the number of lawns from a sample

of 500 Birmingham, Alabama, homes that have various types of warm-season
grasses (bahiagrass, bermudagrass, centipedegrass, St Augustinegrass, and zoysia)
in their lawns. The following bar chart of these data (from Problem 7) is provided
in the file HolidayGrassChart. LO 5
Grasses Grown on Birmingham Area Lawns
Type of Grass

Zoysia

Bermudagrass

St.Augustinegrass
HolidayGrassChart

Bahiagrass

Centipedegrass

0 50 100 150 200

Number of Homes
166 Chapter 4 Purposeful Use of Color

a. Modify this bar chart, which is provided in the file HolidayGrassChart, so that it
uses color more effectively. Also declutter the chart where possible. Explain why
the changes you made produce a chart that is superior to the original bar chart in the
file HolidayGrassChart.
b. Modify bar chart in part a to highlight Bermudagrass.
10. T-Shirt Sales over Time. You have been asked to prepare a line chart that shows
monthly sales for the past year for the four colors of slub cotton t-shirts that are manu-
factured and marketed by the Great Plains Garment Company (GPGC). GPGC sells all
GPGCChart
of its products online. You obtain the data from GPGC’s sales department and produce
the following chart in Excel.
GPGC Monthly Slub T-Shirt Sales
Sales (units)
5000

4000
Azure

3000 Sea Foam

Fuchsia
2000 Plum

1000

0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Upon reviewing your chart, you realize that you could make it more effective
through appropriate use of color symbolism. After deciding to use the color of each
t-shirt for its line on the chart, you obtain the following image showing the four colors
of GPGC’s slub t-shirts from its website collection. The data, the original chart, and the
following image (which can be cut out of the Excel spreadsheet and pasted into Power-
Point) are included in the file GPGCChart. LO 2

a. Use PowerPoint and the Eyedropper tool to determine the hue, saturation, and lumi-
nance of the predominant color of each t-shirt. List the values of the hue, saturation,
and luminance of each color.
b. Revise the line chart in the file GPGC so the color of each line corresponds to the
color of the associated slub t-shirt from part a. Also change the color of the text
in the label at the end of each line to match the color of the corresponding slub
t-shirt.
11. The Relationship between Wins and Attendance in Baseball. In doing an anal-
ysis of the economics of baseball in your state, you collect data on annual home
game attendance and number of wins from 2000–2019 for the state’s two baseball
AttendWinsChart
teams, the Komodos and the Condors. To communicate the nature of the relation-
ship between annual home game attendance and number of wins for the two teams
over that period, you produce the following scatter chart in Excel.
 Problems 167

The chart shows a mildly positive relationship between annual home game attendance
and number of wins for each team. The chart also shows the Condors have routinely won
more games than the Komodos but have generally had lower attendance. However, you
see a way to improve this chart by using the predominant color of each team’s uniform
instead of the Excel default colors for the points that represent the two teams.
You obtain the following images of the two team’s home jerseys from their respec-
tive websites and paste them into the file AttendWinsChart that also contains the data
and your original chart. LO 4

a. Use PowerPoint and the Eyedropper tool to determine the hue, saturation, and lu-
minance of the predominant color of each jersey. List the values of the hue, satura-
tion, and luminance of each color. Images of the two jerseys are included in the file
AttendWinsChart and can be cut and pasted into PowerPoint.
b. Revise the scatter chart in the file AttendWinsChart so the color of each line corre-
sponds to the color of the corresponding t-shirt from part a.
12. Composition of Boise Regional Medical Center Nursing Staff. Boise Regional
Medical Center (BRMC) is completing its annual report for the Idaho Board of Trust-
ees. One section of this report is devoted to an analysis of the composition of the
BRMCChart
BRMC nursing staff. The current draft of this report includes the following bar chart
that shows the percentage of the nursing staff by classification. This chart is included
in the file BRMCChart.
168 Chapter 4 Purposeful Use of Color

Composition of Boise Regional Medical Center

Nursing Staff by Classification

Registered Nurse
Licensed Practical Nurse
Nurse Practitioner
Emergency Room Nurse
Intensive Care Unit Nurse
Medical-Surgical Nurse
Post-Anesthesia Care Nurse
Perioperative Nurse
Telemetry Nurse
0 5 10 15 20 25 30
Percentage of Total Nursing Staff

In its review of the draft report, management has noted that some audience mem-
bers may consider the color of the bars in this chart to be inappropriate for BRMC
because red is often associated with anxiety. Management has requested that the
chart be revised using a color that is more calming. Revise this chart using a color for
the bars that is generally considered to be comforting and calming. Refer to
Figure 4.13 to identify the hue you will use. LO 2
13. Relationship between University Enrollment and Changes in Mean Faculty
Salary. The Ohio Board of Regents (OBR) is preparing its five-year report on
the state of higher education in Ohio. The current draft of the report includes a
OHHigherEdChart
scatter chart of enrollment and the percentage five-year change in average faculty
salary for the 16 largest four-year institutions in the state. The scatter chart that
is included in the current draft of the report follows and is provided in the file
OHHigherEdChart.
 Problems 169

On reviewing the current draft, one of the members of the OBR has asked that
you add an indication of which of these universities is private to this scatter chart.
Use color to distinguish the private universities (Case Western Reserve University,
University of Dayton, Xavier University, Franklin University, University of Findlay)
on the scatter chart from the public universities. What does this additional informa-
tion tell you? LO 3
14. Viewership of Top Rated Scripted Television Shows. The 18–49 age group is an
extremely lucrative demographic for television networks. Advertisers pay a premium to
reach consumers in this age group with their advertising. The following column chart
shows the number of viewers in millions that the top 10 scripted series averaged during
the 2019–2020 television season.

Number of 18–49 Year Old Viewers for

Scripted Television Series
TV Series
This Is Us
Grey’s Anatomy
Chicago PD
Chicago Fire
Modern Family
The Good Doctor
Chicago Med
NCIS
New Amsterdam
Young Sheldon
0 1 2 3 4
Number of Viewers (millions)

It is also important for this graph to show which television network (ABC, CBS,
FOX, NBC) broadcasts each show. The networks that broadcast these shows are:
Top10TV2020Chart ABC: The Good Doctor, Modern Family, Grey’s Anatomy
CBS: Young Sheldon, NCIS
NBC: New Amsterdam, Chicago Med, Chicago Fire, Chicago PD, This Is Us
Use color to add this information to the original bar chart, which is provided in the
file Top10TV2020Chart. Will your chart be difficult for colorblind audience members
to visually process? Explain. LO 3
15. Relationship between Estimated and Actual Time to Complete a Housepainting
Job. Charlene & Daughters (C&D) employees 20 crews of interior house painters that
work throughout the northwest United States. The leader of each crew is responsible
C&DChart
for visiting potential job sites and developing an estimate of the number of hours it will
take to complete a job based on a C&D formula that considers the square footage of
the surfaces to be painted, the amount and type of trim, and whether ceilings are to be
painted. The crew leader also has the latitude to make subjective adjustments based on
other factors that are not considered by the formula.
170 Chapter 4 Purposeful Use of Color

C&D is now assessing the accuracy of its estimates of the time it will take to com-
plete jobs. It has collected the estimated and actual times to complete each of the 10
most recent jobs for each crew, and it has entered these data into the file C&DChart. A
portion of the data, which includes the name of the crew leader, the estimated time to
complete the job, and the actual time to complete the job, follows.

Repair Person Estimated Hours Actual Hours

Briggs 25.8 29.4
Briggs 62.5 66.3
Briggs 75.6 50.9
Briggs 75.7 77.0
Briggs 61.9 74.8
Briggs 6.1 5.8
Briggs 38.2 36.3
Briggs 52.2 55.3
Briggs 63.1 69.3
Briggs 66.9 77.5
Cobb 37.1 32.6
Cobb 45.0 46.6

Taubee 6.8 6.0

Taubee 54.1 62.5
Vincent 25.7 20.3
Vincent 80.9 93.6
Vincent 20.8 20.6
Vincent 24.8 25.7
Vincent 74.4 84.7
Vincent 18.3 18.3
Vincent 74.6 81.2
Vincent 58.7 52.9
Vincent 37.5 41.1
Vincent 68.7 56.0

C&D has also used these data to produce a scatter chart, and it has used color to
differentiate between the crews on this chart. The chart, which follows, is also included
in the file C&DChart. Note that the legend identifies each crew by the name of its crew
leader. LO 3
 Problems 171

Actual Hours
120

Briggs Cobb

Durban Eams

80 Focault Greer

Hale Irwin

Jacks Knoop

Lee Marks

Neal Oliver
40
Perry Quincy

Rand Stratton

Taubee Vincent

0
0 40 80 120
Estimated Hours

a. C&D would like to describe the accuracy of its estimates using a chart. If C&D’s
estimated hours matched the actual hours perfectly, then all points in this scatter
chart would lie on a diagonal line that is 45° from the origin. Draw a diagonal line
at 45° from the origin through the chart to help you assess the accuracy of C&D’s
estimates. To draw this 45° line, draw line that pass through the points (0,0);
(40,40); (80, 80); and (120, 120) on the chart. Would you advise that C&D incorpo-
rate this line into its scatter chart? Why or why not?
b. Revise the chart so all of the points are blue (Hue = 148, Sat = 151, Lum = 152).
Is the use of color in this chart more effective than the use of color in the chart in
part a? Why or why not?
c. C&D has indicated some concern over recent estimates of the crew headed by
Cobb. Use color on the scatter chart to draw attention to the jobs completed by the
crew headed by Cobb and interpret the results.
16. Production of Laptop Computer Screens. CrystalClear, Inc. manufactures low-resolution
15-inch screens for inexpensive laptop computers at plants in five states (California,
Nebraska, North Carolina, North Dakota, and Texas). The number of units produced by
CrystalClear
each plant is provided in the following table and is provided in the file CrystalClear.

Plant Location Units Produced

Texas 54,210
North Dakota 61,002
California 75,143
Nebraska 79,210
North Carolina 82,157
172 Chapter 4 Purposeful Use of Color

CrystalClear, Inc. management suspects that its larger plants are producing the
greatest number of units. The square footage for each plant follows and is also pro-
vided in the file CrystalClear.

Plant Location Sq Ft
Texas 47,000
North Dakota 53,000
California 65,000
Nebraska 63,000
North Carolina 70,000

Provide a scatter chart of the plant data to provide insight into the nature of the rela-
tionship between units produced and size of the plant for the five plants and highlight
the Nebraska plant. LO 4
17. Pet Food Sales by State. Pet Fare, Inc. produces high-end refrigerated pet food
and ships it anywhere in the United States. Total Pet Fare sales of cat food and
dog food by state for the previous year are given in separate worksheets in the file
Cats&Dogs
Cats&Dogs. LO 5
a. Create a geographic information system (GIS) map that uses a sequential
color scheme to represent total Pet Fare cat food sales by state. Use orange for
the color for this choropleth map. What information does this chart convey to
the audience?
b. Create a geographic information system (GIS) map that uses a sequential color
scheme to represent total Pet Fare dog food sales by state. Use blue for the
color for this choropleth map. What information does this chart convey to the
audience?
c. Create a scatter chart that enables the audience to better understand the relationship
between total cat food sales and total dog food sales for the states. What does this
chart tell you about the relationship between total cat food sales and total dog food
sales?
d. Use color to highlight Wyoming in the scatter chart from part c. What does this
chart tell you about total cat food sales and total dog food sales in Wyoming relative
to the other states?
e. What inherent weaknesses do you see in the analysis performed in parts a–c? How
would you address these issues? (Hint: Consider the disparities in the sizes of the
statewide markets.)
18. Alabama Unemployment Rate by County. The Labor Market Information Divi-
sion of the Alabama Department of Labor is preparing its monthly report on unem-
ployment, and it would like to include a choropleth map that shows unemployment
AlabamaUnemployment
rate by county. Use the unemployment rate for each of Alabama’s 77 counties
provided in the file AlabamaUnemployment to create this choropleth map. What
does this map tell you about where the highest rates of unemployment are in
­Alabama? LO 5
19. Life Expectancy of LED Bulbs. Capital Electrics (CE) advertises that the life expec-
tancy of the LED bulbs it produces is 28,000 hours. The electrical engineering team at
CE has designed a test to assess the impact of ambient temperature on the LED light
LEDBulbs
bulbs CE manufactures. Using strict temperature controls, the engineers set up rooms
starting at 0° Fahrenheit in 10-degree increments to 160° Fahrenheit. They then put a
bulb at each wattage from 7 watts to 25 watts in a light fixture in each room, turn on
the fixtures, and record the number of hours until each bulb fails. The data they have
collected are provided in the file LEDBulbs.
 Problems 173

Use these data to create a heat map of the life of the LED bulbs by their wattage and
the ambient temperature of the room in which they are placed. In the New Formatting
Rule dialog box of Conditional Formatting, enter 3-Color Scale for Format Style,
enter Number in the Type: box of the Midpoint column, enter 28000 in the Value:
box of the Midpoint column, and in the Color: row, use orange (Hue = 17,
Sat = 255, Lum = 115) for values below 28,000, white for the Midpoint, and blue
(Hue = 156, Sat = 255, Lum = 48) for values above 28,000. Mask the cell values
and create a legend that tells the audience that orange corresponds to low bulb lives
less than 28,000 hours, blue corresponds to bulb lives greater than 28,000 hours, and
the color grows darker as the deviation of the bulb life from 28,000 increases in either
direction. LO 5
20. Comparing Regional Market Shares. Bay Side Manufacturing produces small
leather goods such as wallets, travel organizers, toiletry kits, briefcases, and folios.
The company has a 16% share in the U.S. travel organizers market, but management
BaySide
is concerned that its market share in the southeastern United States lags behind its
national share. Bay Side’s monthly share of the travel organizers market for the past
year in 15 major cities in the southeastern United States is provided in the file BaySide.
A portion of the data is shown in the following table.

Market Share
Jan Feb Mar Apr Nov Dec
Atlanta 0.142 0.145 0.137 0.152 … 0.132 0.152
Baltimore 0.172 0.199 0.177 0.189 … 0.181 0.185
Baton Rouge 0.133 0.084 0.137 0.123 … 0.144 0.097
Birmingham 0.087 0.105 0.105 0.096 … 0.118 0.121
Charlotte 0.088 0.112 0.094 0.100 … 0.069 0.077
Jacksonville 0.188 0.188 0.190 0.173 … 0.197 0.185
Louisville 0.217 0.214 0.164 0.190 … 0.189 0.192
Memphis 0.107 0.115 0.132 0.111 … 0.125 0.115
Miami 0.162 0.158 0.158 0.157 … 0.180 0.169
Nashville 0.053 0.054 0.074 0.081 … 0.083 0.099
New Orleans 0.129 0.161 0.159 0.189 … 0.194 0.155
Raleigh 0.156 0.188 0.162 0.150 … 0.190 0.191
Richmond 0.178 0.156 0.183 0.157 … 0.158 0.145
Tampa 0.119 0.115 0.108 0.145 … 0.144 0.110
Virginia Beach 0.172 0.174 0.129 0.166 … 0.164 0.191
Washington 0.232 0.241 0.204 0.191 … 0.207 0.243

Use these data to create a heat map of Bay Side’s monthly share of the travel
organizers market for these 15 southeastern U.S. cities. In the New Formatting Rule
dialog box of Conditional Formatting, enter 3-Color Scale for Format Style, enter
Number in the Type: box of the Midpoint column, enter 0.16 in the Value: box of
the Midpoint column, and in the Color: row, use orange (Hue = 17, Sat = 214,
Lum = 143) for values below 0.16 and blue (Hue = 156, Sat = 255, Lum = 48) for
values above 0.16. Mask the cell values and create a legend that tells the audience
that orange corresponds to market shares less than 14%, blue corresponds to market
shares greater than 14%, and the color grows darker as the deviation of the market
share from 14% increases in either direction.
What message will this map communicate to an audience? Will this chart be difficult
for colorblind audience members to visually process? Explain. LO 5
Chapter 5
Visualizing Variability
CONTENTS

 DATA VISUALIZATION MAKEOVER 5-3 UNCERTAINTY IN SAMPLE STATISTICS

Age Distribution of U.S. Gymnasts Displaying a Confidence Interval on a Mean
5-1 CREATING DISTRIBUTIONS FROM DATA Displaying a Confidence Interval on a Proportion
Frequency Distributions for Categorical Data
Relative Frequency and Percent Frequency 5-4 UNCERTAINTY IN PREDICTIVE MODELS
Visualizing Distributions of Quantitative Data Illustrating Prediction Intervals for
a Simple Linear Regression Model
5-2 STATISTICAL ANALYSIS OF DISTRIBUTIONS Illustrating Prediction Intervals for
OF QUANTITATIVE VARIABLES a Time Series Model
Measures of Location SUMMARY
Measures of Variability
GLOSSARY
Box and Whisker Charts
PROBLEMS

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Create and interpret charts used to visualize the LO 5 Describe basic statistical measures of central
frequency distribution of a categorical variable location, variability, and distribution shape
LO 2 Create and interpret histograms and frequency LO 6 Create and interpret a box and whisker chart
polygons, charts used to visualize the distribution
LO 7 Create and interpret visualizations that depict the
of a quantitative variable
uncertainty resulting from sampling error
LO 3 Create and interpret visualizations comparing the
LO 8 Create and interpret charts that depict the uncer-
distributions of two or more variables
tainty in predictions from simple regression mod-
LO 4 Create and interpret strip charts, recognize els and time series models
situations in which to use them, and employ
techniques to improve their clarity
 Data Visualization Makeover 175

D ata V is u ali z ation M a k eover

Age Distribution of U.S. Gymnasts

An article by The Washington Post examined age range for male gymnasts to be 17 years to 30 years
the age range of U.S. Olympic athletes in recent by considering the left end of the purple bar and right
Summer Games. To summarize the analysis, The end of the blue bar. However, there is no information
Washington Post used a chart similar to what is conveyed in this chart to show how male and female
shown in Figure 5.1 for U.S Olympic gymnasts. This gymnasts are distributed over their respective ranges.
chart has a visual similarity to a stacked bar chart, but We can find several opportunities for improving
unfortunately this similarity contributes to confusion the design of the chart in Figure 5.1. We can ease the
when interpreting the chart. In a stacked bar chart, cognitive load and communicate more information
different colors correspond to different quantities, about the age distributions of U.S. Olympic athletes
and an increasing length of a colored bar segment if we use a different type of chart. A chart known as
corresponds to an increase in the number or propor- a frequency polygon, which will be discussed in this
tion of a respective quantity. However, in Figure 5.1, chapter, more effectively communicates insights from
the length of a bar does not convey the same infor- these data to the audience.
mation as a stacked bar chart. Figure 5.1 depicts an Figure 5.2 displays a pair of frequency polygons,
overlapping range bar chart. In a range bar chart, one for male gymnasts and one for female gymnasts.
the endpoints correspond to the smallest and largest In addition to the age range of female and male gym-
values of a variable. Thus, in Figure 5.1 the smallest nasts, the frequency polygons displayed in Figure 5.2
value is 15 and the largest value is 30. provide other information on the age distributions
After referring to the chart legend, we realize that of male and female gymnasts. The line representing
two of the colors correspond to male and female ath- female gymnasts is highest for younger ages and the
letes, respectively, while the third color depicts when line representing male gymnasts is highest for older
the age ranges of male and female athletes overlap. ages, which indicates that the age distribution for
That is, the third color does not correspond to a third female gymnasts is more heavily weighted toward
quantity, but it must be used in conjunction with the younger ages while the age distribution for male
colors of the male and female athletes to make appro- gymnasts is weighted toward older ages. The line for
priate conclusions about the age ranges. This use of female gymnasts peaks at 16 while the peak for male
color is not intuitive and increases the cognitive load of gymnasts peaks at 26, indicating the most common
the audience. In this case, one can determine the age age for a female gymnast is 16 and the most common
range for female gymnasts to be 15 years to 26 years age for a male gymnast is 26. Furthermore, the use of
by considering the left end of the pink bar and right color in Figure 5.2 is straightforward and clearly distin-
end of the purple bar. Similarly, one can determine the guishes between male and female gymnasts.

Figure 5.1 Overlapping Range Bar Chart for Ages of U.S. Olympic Gymnasts
Male Female Both
Age Range of U.S. Gymnasts in the Four Most Recent Summer Games
Age 15 20 25 30 35 40 45 50 55 60

(Continued)
176 Chapter 5 Visualizing Variability

Figure 5.2 Frequency Polygon for U.S. Olympic Gymnasts

Age Distribution of U.S. Olympic Gymnasts
Count of Gymnasts
40
Male Female
35
30
25
20
15
10
5
0
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Age (years)

In this chapter, we address how to visualize the deviations in observed values of data.
These deviations may occur among the values of a variable, such as the age of women
Olympic gymnasts discussed in the Data Visualization Makeover. We describe different
types of charts used to display the distribution of a variable’s values and discuss how the
type of data (categorical or quantitative), the amount of data, and the number of variables
to compare affects the choice of visualization. As an aid in interpreting charts displaying
a variable’s distribution, we define and describe some basic statistical measures of loca-
tion and variability. We conclude the chapter with discussions on how to visually convey
the variability in sample statistics and prediction estimates.

5-1 Creating Distributions from Data

Practically every challenge an organization or individual faces is concerned with the
impact of the possible values of relevant variables will have on an outcome of interest.
Thus, we are concerned with how the value of a variable can vary; variation is the
difference in a variable measured over observations (time, customers, items, etc.). This
variation is often uncertain; we do not perfectly know its magnitude or timing due to
factors beyond our control. In general, a quantity whose values are not known with cer-
tainty is called a random variable.
When we collect data, we are gathering past observed values, or realizations, of a ran-
dom variable. The role of descriptive analytics is to analyze and visualize data to gain a
better understanding of variation and its impact. The frequency distribution of a variable
describes which values were observed and how often those values appear in the data being
analyzed. A frequency distribution can be created for both a categorical variable and a
quantitative variable. For a categorical variable, data consist of labels or names which
cannot be arithmetically manipulated. For a quantitative variable, data consist of numeri-
cal values which can be arithmetically manipulated.
In most cases, it is not feasible to collect data from the entire population of all elements
of interest. In such instances, we collect data from a subset of the population known as a
sample. In the analysis of this chapter, we assume we are dealing with a sample of data rep-
resentative of the population so that generalizations about the entire population can be made.

Frequency Distributions for Categorical Data

It is often useful to visualize a frequency distribution for a categorical variable. A
frequency distribution is a summary of data that shows the number (frequency) of
 5-1 Creating Distributions from Data 177

FIGURE 5.3 Portion of Data in File pop

Pop

observations in each of several nonoverlapping classes, typically referred to as bins.

Consider 50 soft drink purchases. Figure 5.3 displays a portion of these data. Each pur-
chase is for one of five popular soft drinks, which define the five bins: Coca-Cola, Diet
Coke, Dr. Pepper, Pepsi, and Sprite.
We can use Excel to calculate the frequency of categorical observations occurring in a
data set and then display the frequency distribution. The following steps show how to cre-
ate a column chart in Excel using the data in the file Pop.
Step 1. Select cells A1:A51
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Recommended Charts button Recommended
Charts
in the Charts group
Step 4. When the Insert Chart dialog box appears:
Select Clustered Column
Click OK
Step 5. Click any of the columns in the chart that appears. While the columns are
selected, right click a column, then select Sort and Sort Largest to Smallest
The chart in Figure 5.4 Figure 5.4 displays the output of Steps 1 through 5, which generate a PivotTable to sum-
(and all figures of this chapter) marize the data with a frequency distribution and a PivotChart to visualize the frequency dis-
have been edited using steps
tribution. As Figure 5.4 shows, Coca-Cola appears 190 times, Pepsi appears 130 times, Diet
described in Chapters 2 and 3.
Coke appears 80 times, Sprite appears 50 times, and Dr. Pepper appears 50 times. This fre-
quency distribution provides a summary of how the 500 soft drink purchases are distributed
We discuss PivotTables and
across the five soft drinks. The frequency distribution shows that Coca Cola is the leader,
PivotCharts in more detail in Pepsi is second, Diet Coke is third, and Sprite and Dr. Pepper are tied for fourth. The fre-
Chapter 6. quency distribution thus summarizes information about the popularity of the five soft drinks.
In Step 5, we ordered the data from largest to smallest to facilitate the comparison of the
different soft drink types. This reordering of the data is acceptable because there is no ordi-
nal relationship between soft drinks. However, if an ordinal relationship exists between bin
categories, then it would not be appropriate to sort the bins by the number of observations
in each bin. Instead, the bins should be ordered according to their ordinal relationship.
If an analyst prefers not to use the Recommended Charts approach to construct a
PivotTable and PivotChart to display a frequency distribution for categorical data, Excel func-
tions and a column chart can be used to do so in a more manual fashion as we explain next.
Referring again to the data in the Pop file, we count the number of times each soft
drink appears using the COUNTIF function. Figure 5.5 shows a portion of the 500 soft
drink purchases. In the cells C2 through C6, we enter the five different soft drinks
as the bin labels. In cell D2, we enter the formula 5COUNTIF(A:A, C2), where A:A
refers to the column range containing the data, and C2 is the bin label (Coca-Cola) that
we are trying to match. The COUNTIF function in Excel counts the number of times a
178 Chapter 5 Visualizing Variability

FIGURE 5.4 PivotTable and PivotChart of Frequency Distribution of Soft Drink Purchase Data

FIGURE 5.5 Creating a Frequency Distribution for Categorical Data Using COUNTIF Function
 5-1 Creating Distributions from Data 179

certain value appears in the indicated range. In this case we want to count the number
of times Coca-Cola appears in the data. The result is a value of 190 in cell D2, indicat-
ing that Coca-Cola appears 190 times in the data. We can copy the formula from cell
Chapter 2 outlines the steps
D2 to cells D3 through D6 to get frequency counts for Diet Coke, Dr. Pepper, Pepsi,
for constructing a clustered and Sprite. Using the data in C2:D6, a clustered column chart can then be used to illus-
column chart. trate the distribution of soft drink purchases.

Relative Frequency and Percent Frequency

A frequency distribution, such as the one in Figure 5.4, shows the number (count) of items
in each of several nonoverlapping bins. However, we are often interested in the proportion,
or percentage, of items in each bin. The relative frequency of a bin equals the fraction or
proportion of items belonging to a class. For a data set with n observations, the relative fre-
quency of each bin can be determined as follows:

Frequency of the bin

     Relative frequency of a bin 5  (5.1)
n
Often a relative frequency is expressed as a percentage. The percent frequency of a bin is
the relative frequency multiplied by 100. To obtain a percent frequency distribution for the
data in the Pop file, we continue from the PivotTable in Figure 5.4 with the following steps:
Step 1. Select any cell in the Count of Soft Drink Purchase column of the PivotTable
(any cell in range B3:B8)
Step 2. When the PivotTable Fields task pane appears:
In the Values area, select the triangle to the right of Count of Soft Drink
Purchase Count of Soft Drink Purchase
Select Value Field Settings… from the list of options.
Step 3. When the Value Field Settings dialog box appears:
Click the Show Values As tab and in the box below Show values as
select % of Grand Total
A percent frequency Figure 5.6 shows the result of the preceding steps. The percent frequency for Coca-Cola is
distribution can be 190/500 5 0.38 5 38%, the percent frequency for Pepsi is 130/500 5 0.26 5 16%, and so on.
manually constructed in the
We can also note that 38% 1 26% 1 16% 5 80% of the purchases were the top three soft drinks.
spreadsheet in Figure 5.5 by
implementing Equation (5.1)
A percent frequency distribution can be used to provide estimates of the relative like-
as a spreadsheet formula in lihoods of different values for a random variable. So, by constructing a percent frequency
the range E2:E6, respectively. distribution from observations of a random variable, we can estimate the probability
distribution that characterizes its variability. For example, suppose that a concession stand
has determined it will procure a total of 12,000 ounces of soft drinks for an upcoming
concert, but it is uncertain how to divide this total over the individual soft drink types.
However, if the data in the Pop file are representative of the concession stand’s customer
population, the manager can use this information to determine appropriate volumes of
each type of soft drink. For example, the data suggest that the manager should procure
12,000 3 0.38 5 4,560 ounces of Coca-Cola.
A prominent example of a relative frequency distribution often used in accounting
and finance is Benford’s Law, which states that in many data sets, the proportion of
observations in which the first digit is 1, 2, 3, 4, 5, 6, 7, 8, or 9, respectively, follows
the distribution in Figure 5.7. Benford’s Law applies to a variety of naturally occur-
ring data sets, including item prices, utility bills, street addresses, corporate expense
reports, city populations, and river lengths. It tends to be most applicable in data sets
governed by a power law, in which a variable of interest experiences a proportional
change in response to a change in one or more other variables. When dealing with data
in an application that typically obeys Benford’s Law, if the relative frequency distribu-
tion of the first digits of these data deviates substantially from the relative frequency
distribution in Figure 5.7, then there may be systemic errors in the data or the data may
be fraudulent.
180 Chapter 5 Visualizing Variability

FIGURE 5.6 Percent Frequency Distribution of Soft Drink Purchase Data

FIGURE 5.7 Relative Frequency Distribution of Data Obeying Benford’s Law

Benford’s Law
Proportion of First-Digit Observations
30.1%

17.6%
12.5%
9.7%
7.9% 6.7% 5.8% 5.1% 4.6%

1 2 3 4 5 6 7 8 9
First Digit

Figure 5.7 indicates that, according to Benford’s Law, approximately 30% of the values
in an applicable data set would begin with a digit of 1, while only about 4.6% of the values
would begin with a digit of 9. This is different than if we assumed that the first digits of the
values would all be equally likely, in which case we would expect 1/9 < 11.1% of the val-
ues in a data set to start with each digit.
 5-1 Creating Distributions from Data 181

Visualizing Distributions of Quantitative Data

As with categorical data, we can create frequency distributions for quantitative data, but we
must be more careful in defining the nonoverlapping bins to be used in the frequency dis-
tribution. Recall that for categorical data, a frequency distribution’s bins are based on the
different categories. For quantitative data, each bin in the frequency distribution is based on
the range of values that the bin contains.
To create a frequency distribution for quantitative data, three features need to be defined:
1. The number of nonoverlapping bins
2. The width (numerical range) of each bin
3. The range spanned by the set of bins
Excel possesses functionality that automatically defines each of these features. To demon-
strate, consider a data set that contains the age at death for 700 individuals. Figure 5.8 dis-
plays a portion of this data, contained in the file Death. The following steps construct the
histogram in Figure 5.9 illustrating the distribution of the ages at death.
Step 1. Select cells A1:A701
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Statistic Chart button in the Charts group
When the list of statistic charts appears, select Histogram

FIGURE 5.8 Portion of Data in File Death

Death

As Figure 5.9 illustrates, a histogram is simply a column chart with no spaces between
the columns whose heights represent the frequencies of the corresponding bins. Eliminat-
ing the space between the columns allows a histogram to reflect the continuous nature of
the variable of interest. For this data set, Excel automatically chose to use 16 bins, each
spanning 7 years, traversing the range from 0 to 112.
Returning to Figure 5.9, we observe that the tallest column corresponds to the bin (77, 84].
To interpret this bin, we note that a square bracket indicates that the end value is included in
the bin, and a round parenthesis indicates the end value is excluded. So the most common
ages at death occur in the range greater than 77 years old and less than or equal to 84 years
old. Furthermore, we observe the data are highly skewed left with most individuals dying at
relatively old ages, but small numbers of individuals dying at young ages.
The choice of number of bins and bin width can strongly affect a histogram’s display
of a distribution. If an analyst prefers not to use the automatic histogram generated by
Excel’s Charts functionality, more user control is possible by using the Excel func-
tion FREQUENCY and a column chart to construct a histogram. Again, we will use
the 700 observations in the file Death to manually create a histogram. The first step in
182 Chapter 5 Visualizing Variability

FIGURE 5.9 Histogram for the Age at Death Data

Distribution of Age at Death

Count of People
160
140
120
100
80
60
40
20
0

(9 1]
(1 4]
(2 1]
(2 8]
(3 5]
(4 2]
(4 9]
(5 6]
(6 3]
(7 0)
(7 7]
(8 4]

(9 8]
05 ]

]
(7 ]

(1 105
12
,7

9
,1

2
2
3
4
4
5
6
7
7
8

9
[0

,1
4,
4,
1,
8,
5,
2,
9,
6,
3,
0,
7,

1,
8,
Age at Death (years)

manually creating a histogram is determining the number of bins, the width of the bins,
and the range spanned by the bins.

Number of Bins Bins are formed by specifying the ranges used to group the data. As a
general guideline, we recommend using from 5 to 20 bins. Using too many bins results
in a histogram in which many bins contain only a few observations. With too many bins,
the histogram does not capture generalizable patterns in the distribution and instead may
appear jagged and “noisy.” Using too few bins results in a histogram that aggregates
observations with too wide of range of values into the same bins. With too few bins, the
histogram fails to accurately capture the variation in the data and presents only blurred
high-level patterns. For a small number of observations, as few as five or six bins may be
used to summarize the data. For a larger number of observations, more bins are usually
required. The determination of the number of bins is an inherently subjective decision,
and the notion of a “best” number of bins depends on the subject matter and goal of the
analysis. Because the number of observations in the Death file is relatively large
(n 5 700), we should choose a larger number of bins. We will use 16 bins to match Figure 5.9.

Width of the Bins As a general guideline, we recommend that the width be the same for
each bin. Thus, the choices of the number of bins and the width of bins are not independent
decisions. A larger number of bins means a smaller bin width and vice versa. To determine
an approximate bin width, we begin by identifying the largest and smallest data values.
Then, with the desired number of bins specified, we can use the following expression to
determine the approximate bin width.

largest data value 2 smallest data value

   approximate bin width 5 (5.2)
number of bins

The approximate bin width given by Equation (5.2) can be rounded to a more conve-
nient value based on the preference of the person developing the frequency distribution. For
example, using equation (5.2) supplies an approximate bin width of (109 2 0)/16 5 6.8125.
We round this number up to obtain a bin width of 7. By rounding up, we assure that 16 bins
 5-1 Creating Distributions from Data 183

of width 7, with the first bin starting from the smallest data value, will cover the range of
values in the data.

Range Spanned by Bins Once we have set the number of bins and the bin width, the
remaining decision is to set the value at which the first bin begins. We must ensure that
the set of bins spans the range of the data so that each observation belongs to exactly
one bin. For example, a histogram with 16 bins and a bin width of 7 will cover a range
of 112. Considering the Death data, we observe that the smallest data value is 0 and
the largest data value is 109. Because the range of data is 109, but the range of the
bins is 112, we have four possible choices of the value to begin the first bin that would
result in a set of bins that places each value from the data in a bin. We can define the
first bin as [23, 4] or [22, 5] or [21, 6] or [0, 7]. For example, if we begin the first
bin at the smallest data value, the first bin is [0, 7], the second bin is (7, 14], the third
bin is (14, 21]…, and the 16th bin is (105, 112]. In this case, the 16th bin’s range
extends past the largest data value of 109.
In Figure 5.10, columns C and D contain the lower and upper limits defining the
bins. We count the number of observations falling within each bin’s range using the
In older versions of Excel, you FREQUENCY function. In cell E2, we enter the formula 5FREQUENCY(A2:A701,
must enter the FREQUENCY
D2:D17), where A2:A701 is the range for the data, and D2:D17 is the range containing the
function as an array function
by highlighting cells
upper limits for each bin. After pressing Enter in cell E2, the range E2:E18 is populated
E2:E17 before entering the with the number of observations falling within each bin’s range.
formula and then pressing Using the data in D2:E17, a clustered column chart can then be used to illustrate the dis-
Ctrl1Shift1Enter rather than tribution of ages at death. In column F, we use the Excel CONCAT function to create a set
just Enter.
of bin labels that we will use for the horizontal axis. The CONCAT function simply com-
In older versions of Excel,
bines elements from different cells and/or different text stings into the same cell.
the CONCAT function may
Step 1. Select cells D2:E17
not exist and you must use
the CONCATENATE function
Step 2. Click the Insert tab on the Ribbon
instead. Step 3. Click the Insert Column or Bar Chart button in the Charts group
When the list of column and bar charts subtypes appears, click the
Clustered Column button
Steps 1 through 3 result in a plotting both Bin Upper Limit and Frequency. To correct this,
we need the following steps:
Step 4. Right click the chart and select Change Chart Type…
Step 5. When the Change Chart Type task pane appears, select the Cluster Column
type that plots the appropriate number of variables (in this case, the single vari-
able Frequency plotted with 16 columns ) and click OK

Step 6. Right click one of the columns in the chart and select Format Data Series…

When the Format Data Series task pane opens, click the Series Options

button and set the Gap Width to 0%

Click the Fill &Line button , under Border select Solid line, and to
the right of Color, select white from the drop down menu.
Step 7. Right click the chart and select Select Data…
When the Select Data Source dialog box appears, click Edit under
Horizontal (Category) Axis Labels. In the Axis Labels dialog box,
enter 5Data!$F$2$F$17 in the box below Axis label range: and
click OK
Click OK to close the Select Data Source dialog box
184 Chapter 5 Visualizing Variability

Further editing will result in a histogram that matches Figure 5.9.

FIGURE 5.10 Creating a Frequency Distribution for Quantitative Data

Using the Excel FREQUENCY Function

DeathFrequency

We note that the choice of the number of bins (and the corresponding bin width) may
change the shape of the histogram (particularly for small data sets). Therefore, it is com-
mon to determine the number of bins and the appropriate bin width by trial and error. Once
a possible number of bins are chosen, equation (5.2) is used to find the approximate bin
width. The process can be repeated for several different numbers of bins.
To illustrate the impact of varying the number of bins and bin width on the shape of
the histogram, we consider 30 observations in the Death30 file, a subset of the data in
the Death file. Figure 5.11 depicts three histograms, one using 5 bins with width 12,
one using 8 bins with width 8, and one using 10 bins with width 6. As Figure 5.11
illustrates, the choice of binning parameters can affect the shape of the histogram’s
distribution. The histogram with 5 bins and the histogram with 10 bins both suggest
that the bin containing the oldest age range is the most likely. However, the histogram
with 8 bins suggests that the bin containing the second oldest age range is most likely
and the oldest age range is the third-most likely. An inspection of the data reveals that
6 out of the 30 observations had an age of death of 87 years, making the display of the
distribution highly sensitive to the bin that contains 87.
One of the most important uses of a histogram is to provide information about the
shape, or form, of a distribution. Skewness, or the lack of symmetry, is an important char-
acteristic of the shape of a distribution. Figure 5.12 contains four histograms constructed
from relative frequency distributions that exhibit different patterns of skewness. Figure 5.12a
shows the histogram for a set of data moderately skewed to the left. A histogram is said to
be skewed to the left if its tail extends farther to the left than to the right. This histogram
is typical for exam scores, with no scores above 100%, most of the scores above 70%, and
only a few really low scores.
 5-1 Creating Distributions from Data 185

FIGURE 5.11 Effect of Varying Number of Bins in a Histogram

Distribution of Age at Death: 5 Bins

Count of People
16
14
12
10
8
6
4
2
0
[31, 43] (43, 55] (55, 67] (67, 79] (79, 91]
Age at Death (years)
Distribution of Age at Death: 8 Bins
Count of People
16
14
12
10
8
Death30
6
4
2
0
[31, 39] (39, 47] (47, 55] (55, 63] (63, 71] (71, 79] (79, 87] (87, 95]
Age at Death (years)

Distribution of Age at Death: 10 Bins

Count of People
16
14
12
10
8
6
4
2
0
[31, 37] (37, 43] (43, 49] (49, 55] (55, 61](61, 67] (67, 73] (73, 79] (79, 85] (85, 91]
Age at Death (years)

Figure 5.12b shows the histogram for a set of data moderately skewed to the right. A
histogram is said to be skewed to the right if its tail extends farther to the right than to the
left. An example of this type of histogram would be for data such as housing prices; a few
expensive houses create the skewness in the right tail.
Figure 5.12c shows a symmetric histogram, in which the left tail mirrors the shape of
the right tail. Histograms for data found in applications are rarely perfectly symmetric,
but the histograms for many applications are roughly symmetric. Data for SAT scores, the
heights and weights of people, and so on lead to histograms that are roughly symmetric.
186 Chapter 5 Visualizing Variability

Figure 5.12d shows a histogram highly skewed to the right. This histogram was con-
structed from data on the amount of customer purchases in one day at an apparel store.
Data from applications in business and economics often lead to histograms that are skewed
to the right. For instance, data on wealth, salaries, purchase amounts, and so on often result
in histograms skewed to the right.

FIGURE 5.12 Histograms Showing Distributions with Different Levels

of Skewness

Moderately Skewed Left Moderately Skewed Right

Relative Frequency Relative Frequency
0.35 0.35
0.3 0.3
0.25 0.25
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
Bins Bins
(a) (b)

Symmetric Highly Skewed Right

Relative Frequency Relative Frequency
0.35 0.35
0.3 0.3
0.25 0.25
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
Bins Bins
(c) (d)

As we have shown, column charts and histograms can be effective ways to visualize the
distribution of a variable. However, when comparing the distribution of two or more vari-
ables, these columnar displays become cluttered. Next, we present a visualization tool that
is particularly helpful for visualizing the distributions of multiple variables.
A frequency polygon is a visualization tool useful for comparing distributions, particularly
for quantitative variables. Like a histogram, a frequency polygon plots frequency counts of obser-
vations in a set of bins. However, a frequency polygon uses lines to connect the counts of differ-
ent bins, in contrast to a histogram, which uses columns to depict the counts in different bins.
To demonstrate the construction of histograms and frequency polygons for two different
variables, we consider the data in the file DeathTwo, which supplements the age at death infor-
mation for the 700 individuals in the file Death with the sex of each of these individuals. Similar
to how we constructed the frequency distribution for all 700 observations, we must create sepa-
rate frequency distributions for the female and male observations, respectively. However, when
comparing frequency distributions, it is a good practice to use relative frequency calculations
because the total number of observations in the two distributions may not be the same. For
instance, in the file DeathTwo, there are 327 female observations and 373 male observations, so
comparing just the count of observations in the bins may distort the comparison.
 5-1 Creating Distributions from Data 187

In Figure 5.13, we compute the number of female and male observations by

entering the formula 5COUNTIF(B2:B701, “Female”) in cell F3 and the formula
5COUNTIF(B2:B701, “Male”) in cell G3. We use these counts to compute the rel-
DeathTwo ative frequencies of male and female observations in the bins, which are defined by
the lower and upper limits in Columns D and E. For the female and male observations
(which have already been sorted), we count the number of observations falling within
each bin’s range using the FREQUENCY function. In cell F6, we enter the formula
5FREQUENCY(A2:A328,E6:E21)/$F$3, where A2:A328 is the range for the female
observations, E6:E21 is the range containing the upper limits for each bin, and cell
F3 contains the total number of female observations. After hitting enter in cell F6, the
range F6:F22 is populated with the relative frequency of female observations falling
within each bin’s range. In cell G6, we enter the formula 5FREQUENCY(A329:A701,
E6:E21)/$G$3, where A329:A701 is the range for the male observations, E6:E21 is the
range containing the upper limits for each bin, and cell G3 contains the total number of
male observations. After hitting enter in cell G6, the range G6:G21 is populated with
the relative frequency of male observations falling within each bin’s range.

FIGURE 5.13 Creating Frequency Distributions for Female and Male

Observations in the File DeathTwo

DeathTwoFrequency

The following steps produce a chart with histograms for the male and female observations.
Step 1. Select cells F6:G21
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Column or Bar Chart button in the Charts group
Step 4. When the list of column and bar chart subtypes appears, click the Clustered
Column button
188 Chapter 5 Visualizing Variability

Further editing will result in the clustered column chart displayed in Figure 5.14a.
To display the age at death distributions for female and male observations as a frequency
polygon, we execute the following steps.
Step 1. Select cells F6:G21
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Line or Area Chart button in the Charts group
When the list of line and area chart subtypes appears, click the Line button
Further editing will result in the frequency polygons displayed in Figure 5.14b.

FIGURE 5.14 Comparing Frequency Polygon Display to Clustered

Column Display

Distribution of Age at Death

Relative Frequency Male Female
0.30
0.25
0.20
0.15
0.10
0.05
0.00
,7]

(2 ]
(2 ]
(3 ]
(4 ]
(4 ]
(5 ]
(6 ]
(7 ]
(7 ]
(8 ]
(9 ]
8]

05 ]
2]
,14

1
8
5
2
9
6
3
0
7
4
1

05
4,2
1,2
8,3
5,4
2,4
9,5
6,6
3,7
0,7
7,8
4,9
1,9

,11
(0

8,1
(7
(1

(9
(1
Age at Death (years)
(a)

Distribution of Age at Death

Relative Frequency
Male Female
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1]

(2 8]
5]

(4 ]

(4 ]
6]

(6 ]
,7]

(7 ]

(7 ]
(8 ]
(9 ]
8]

(1 05]

2]
2
9

3
,14

0
7

4
1
4,2
1,2
8,3

5,4
2,4

9,5

6,6
3,7
0,7

7,8
4,9
1,9

,11
(0

8,1
(7
(1
(2

(3

(5

05
(9

Age at Death (years)

(b)

As Figure 5.14 illustrates for the age at death distributions for males and females, the use
of lines in the frequency polygons preserves the continuity of each distribution’s shape,
while the use of a clustered column chart provides a disrupted visualization of the distribu-
tions. While frequency polygons provide for a more transparent comparison of two or more
distributions, for a single distribution they do not support the magnitude comparison of
different bins as well as a histogram. Therefore, histograms are typically preferred for the
visualization of a single variable’s distribution.
When comparing many distributions (three or more), frequency polygons plotted on the
same chart can become cluttered. For shape comparisons of multiple distributions, an arrange-
ment of individual visualizations in a trellis display can be helpful. A trellis display is a vertical
5-1 Creating Distributions from Data 189

or horizontal arrangement of individual charts of the same type, size, scale, and formatting that
differ only by the data they display. Figure 5.15 contains a vertical trellis display of the length
of stay distributions of three hospitals using frequency polygons. This trellis display facilitates
distribution shape comparisons, but it is not as useful for magnitude comparisons.

FIGURE 5.15 Trellis Display of Length of Stay Distributions for Three

Hospitals

Length of Stay: Seattle Grace

Percent Frequency
50%
40%
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10 11 12
Days

Length of Stay: Seattle Mercy West

Percent Frequency
50%
40%
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10 11 12
Days

Length of Stay: Grey Sloan Memorial

Percent Frequency
50%
40%
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10 11 12
Days

One shortcoming of both histograms and frequency polygons is the specific values of
the smallest and largest values are difficult to discern from the visualization due to the bin-
ning of values. If we want to display a small set of values in a manner that shows the indi-
vidual values, a visualization known as the strip chart can be useful.
To demonstrate a strip chart, consider the data in the file HalfMarathon, which contains
times for a collection of runners in a competitive half-marathon race. Figure 5.16 displays
a portion of these data. The following steps construct horizonal strip charts displaying the
times for the male and female runners, respectively.
190 Chapter 5 Visualizing Variability

FIGURE 5.16 Portion of the data in File HalfMarathon

HalfMarathon

Step 1. Select cells A1:B54

Step 2. Click the Data tab on the Ribbon
Z A
Step 3. Click the Sort button A Z in the Sort & Filter group
Sort
Step 4. When the Sort dialog box appears, select the check box for My data has headers
In the Sort by row, select Sex for the Column entry, Cell Values for Sort
on entry, and Z to A for Order entry
Click Add Level
In the Then by row, select Time (Minutes) for the Column entry, Cell
Values for Sort on entry, and Smallest to Largest for Order entry
Click OK
Step 5. For each cell in the range C2:C23 (each cell corresponding to a Male finishing
time), enter the value 10
Step 6. For each cell in the range C24:C54 (each cell corresponding to a Female fin-
ishing time), enter the value 20
Step 7. Select cells B2:C23 (cells corresponding to Male finishing times)
Step 8. Click the Insert tab on the Ribbon
Step 9. Click the Insert Scatter (X,Y) or Bubble Chart button in the Charts group
Step 10. When the list chart subtypes appears, click the Scatter button
Right click the chart and select Select Data…
Step 11. When the Select Data Source dialog box appears:
Select Series1 under Legend Entries (Series)
Click the Edit button Edit

Step 12. When the Edit Series dialog box appears:

Enter Male in the box under Series name:
Click OK to close the Edit Series dialog box
In a horizontal strip chart, the
Step 13. In the Select Data Source dialog box, click the Add button Add

vertical axis has no meaning. Step 14. When the Edit Series dialog box appears:
In this example, we plot male Enter Female in the box under Series name:
half-marathon times at a Enter 5Data!$B$24:$B$54 in the box under Series X values:
height of 10 and female half-
marathon times at a height of
Enter 5Data!$C$24:$C$54 in the box under Series Y values:
20 to create visual separation, Click OK to close the Edit Series dialog box
but these values are arbitrary. Click OK to close the Select Data Source dialog box
5-1 Creating Distributions from Data 191

After editing, these steps produce the strip chart in Figure 5.17. Figure 5.17 displays
each half-marathon time for males and females, respectively, and we can see the fastest
and slowest times for each sex. However, this strip chart fails to clearly show the relative
density of half-marathon times over the range like a histogram or frequency polygon
because the vertical axis in the strip chart has no meaning. Furthermore, as the number of
values to plot increases and when there are multiple values that are the same or nearly the
same, a strip chart suffers from occlusion. Occlusion is the inability to distinguish some
individual data points because they are hidden behind others with the same or nearly the
same value. Occlusion in strip charts can be mitigated by (1) plotting hollow dots rather
than filled dots and (2) jittering the observation. Jittering an observation involves slightly
adjusting the value of one or more of the variables comprising the observation.

FIGURE 5.17 Strip Chart of Female and Male Half-Marathon Race Times

Half-Marathon Race Times

Male Female

0 20 40 60 80 100 120 140 160 180 200

Minutes

To address the occlusion in Figure 5.17, starting from this chart, we execute the following steps.
Step 15. Right click the Female data series in the chart and select Format Data Series…
Step 16. When the Format Data Series task pane appears:
Click the Fill & Line button
Click Marker Fill
Under Fill, select No fill
Step 17. Leaving the Format Data Series task pane open, click the Male data series in
the chart
Click the Fill & Line button
Click Marker Fill
Under Fill, select No fill
Steps 15–17 plot the half-marathon times using hollow dots. To jitter the observations ver-
tically, we execute steps 18–23 to produce Figure 5.18. Specifically, in steps 18 and 19, we
jitter by adding a small random number between zero and one to the height at which male
and female half-marathon times are plotted.
Step 18. In cell D2, enter the formula 5C21RAND()
Copy the formula in cell D2 to cells D3:D54
Step 19. Right click the chart and choose Select Data…
Step 20. When the Select Data Source dialog box appears:
In the Legend Entries (Series) area, click Male and then click Edit
Step 21. When the Edit Series dialog box, appears:
enter 5Data!$D$2:$D$23 in the box under Series Y values:
Click OK to close the Edit Series dialog box
192 Chapter 5 Visualizing Variability

Step 22. In the Select Data Source dialog box:

In the Legend Entries (Series) area, click Female and then click Edit
Step 23. When the Edit Series dialog box appears:
Enter 5Data!$D$24:$D$54 in the box under Series Y values:
Click OK to close the Edit Series dialog box
Click OK to close the Select Data Source dialog box

FIGURE 5.18 Jittered Strip Chart of Female and Male Half-Marathon

Race Times

Half-Marathon Race Times

Female

Male

0 20 40 60 80 100 120 140 160 180 200

Minutes

Compared to Figure 5.17, the hollowing and jittering of the plotted values in Figure
5.18 results in a strip chart that more clearly display the density of similar half-marathon
times. Recall the vertical axis has no meaning, so adding a small random value between
zero and one to the y-series values does not alter the interpretation of the chart at all, but it
allows the audience to visually discern between similar half-marathon times. If necessary,
we could have also jittered the x-series values by adding and subtracting relatively small
values to the half-marathon times without qualitatively changing the insight derived from
the chart, but that was not necessary in this case.

N o te s 1 C o m m en t s

1. When jittering the data points of a chart, it may be neces- density chart smooths the extremes of the histograms in an
sary to add or subtract values more extreme than random attempt to generalize the patterns in the data. Excel does
values between zero and one. In these cases, the Excel not have built-in functionality to construct a kernel density
formula 5a1RAND()*(b2a) can be used to generate a ran- chart (which is not the same as a frequency polygon), but
dom value between a and b to be added to a data point. many statistical software packages such as R do.
For example, 5251RAND()*(52(25)) generates a random Distribution of Age at Death
value between 25 and 5. Probability Density
2. A kernel density chart is a “continuous” alternative to his- 0.04

tograms designed to overcome the reliance of histograms

0.03
on the choice of number of bins and bin width. Kernel den-
sity charts employ a smoothing technique known as kernel 0.02
density estimation to generate a more robust visualization
of the distribution of a set of values. As an example, a kernel 0.01

density chart for the 30 observations in Death30 file is dis-

0.00
played with this note. Comparing the kernel density chart 20 40 60 80 100

to the histograms in Figure 5.11, we observe that the kernel Age at Death (years)
 5-2 Statistical Analysis of Distributions of Quantitative Variables 193

5-2 Statistical Analysis of Distributions

of Quantitative Variables
In this section, we introduce some basic statistical measures for describing the distribution
of a quantitative variable. We then present data visualizations that utilize these statistical
measures.

Measures of Location
A measure of (central) location identifies a single value of a variable that in some manner
best characterizes the entire set of values. In this sense, a measure of location is a measure
of a variable’s center around which other values are distributed. In this section, we present
different measures of location and discuss their relative advantages and disadvantages.
A common measure of central location is the mean, or average value, for a variable. To
illustrate the computation of the mean of a set of sample values, consider the 12 home sales
in a Cincinnati, Ohio, suburb listed in the CincySales file and displayed in Figure 5.19. The
mean of these 12 values is
456,400 1 298,000 1 Á 1 108,000
5 $219,950
12
The mean of a variable can be found in Excel using the AVERAGE function. In Figure 5.19,
the formula 5AVERAGE(A2:A13) in cell D2 calculates the mean home sale value of $219,950.
The median, another measure of central location, is the value in the middle when the
data are arranged in ascending order (smallest to largest value). With an odd number of
observations, the median is the middle value. An even number of observations has no sin-
gle middle value. In this case, we define the median as the average of the values for the
middle two observations. The median of the 12 Cincinnati home sales is the average of the
sixth and seventh observations as computed by
$208,000 1 $199,500
5 $203,750
2
The median of a variable can be found in Excel using the function MEDIAN. In Figure 5.19,
the formula 5MEDIAN(A2:A13) in cell D3 calculates the median home sale value of $203,750.
Although the mean is a commonly used measure of central location, its calculation is
influenced by outlying values–extremely small and extremely large values. Therefore,
the median is often the preferred measure of central location as its calculation is resistant
to outlying values. Notice that the median is smaller than the mean in Figure 5.19. This
is because the one large value of $456,400 in our data set inflates the mean but does not
affect the median. Notice also that the median would remain unchanged if we replaced the
$456,400 with a sales price of $1.5 million. In this case, the median selling price would
remain $203,750, but the mean would increase to $306,916.67. If you were looking to buy
a home in this suburb, the median gives a better indication of the central selling price of the
homes there. We can generalize, saying that whenever a data set contains extreme values or
is severely skewed, the median is the preferred measure of central location; this is particu-
larly true for data sets with relatively few observations.
A third measure of location, the mode, is the value that occurs most frequently in a data
set. Occasionally the greatest frequency occurs at two or more different values, in which
case more than one mode exists. If no value in the data occurs more than once, we say the
data have no mode. In the Cincinnati home sales data, there are two values that each occur
In older versions of Excel, the
twice and all other values occur just once. Therefore, the two modes are $254,000 and
MODE.MULT function must
be executed by entering
$138,000. All modes of a variable can be found in Excel using the function MODE.MULT.
Ctrl+Shift+Enter rather than In Figure 5.19, the formula 5MODE.MULT(A2:A13) in cell D4 calculates the two modes
just Enter. of $254,000 and $138,000 and places them in cells D4 and D5, respectively.
194 Chapter 5 Visualizing Variability

FIGURE 5.19 Measures of Location for Cincinnati Home Sales Data

CincySales

The mode can be a useful measure of central location for variables that have a relatively
small set of distinct values. For variables with many possible values (such as the value of
home sales in the CincySales file or the race times in the HalfMarathon file), the frequency
that defines the mode will either be small or the mode may not exist. For variables with many
possible values, it may be best to construct a histogram and apply the notion of the mode to
refer to the bin (range of values) with the most observations. That is, the bin in a histogram
with the most observations (the tallest column) may then be referred to as the mode.

Measures of Variability
While measures of location provide a single central value that in some sense is most char-
acteristic for a sample of a variable’s values, these measures fail to convey any informa-
tion regarding the variability in the values. For instance, the median home sale value of
$203,750 for the Cincinnati home sales data provides no information on how spread out the
set of 12 home sale values are. So, in addition to measures of location, it is often desirable
to consider measures of variability, or dispersion.
The simplest measure of variability is the range. The range can be found by subtracting
the smallest value from the largest value in a data set. For the Cincinnati home sales data,
the range is
$456,400 2 $108,000 5 $348,400
5-2 Statistical Analysis of Distributions of Quantitative Variables 195

Excel does not offer a range function, but the range of a variable can be found in Excel
using the MAX and MIN functions. In Figure 5.20, the formula 5MAX(A2:A13) 2
MIN(A2:A13) in cell D7 calculates the range of home sale values to be $348,400.
Although the range communicates a notion of variability by supplying how much
the largest value differs from the smallest value, it is seldom relied up on the only
measure of variability. This is because the range is based on only two of the observa-
tions and thus is highly influenced by extreme values. For example, in the Cincinnati
home sales data, range provides no notion of how close or how far apart the other
10 home sale values are; it only tells us that the largest home sale and smallest home
sale are $348,400 apart.

FIGURE 5.20 Measures of Variability for Cincinnati Home Sales Data

Another common measure of variability is standard deviation, which is based on how

much each observation’s value deviates from the mean. The standard deviation of a sam-
ple of a variable’s values can be viewed as the average amount that an observation in the
sample deviates from the sample mean. For the Cincinnati home sales data, the standard
deviation is computed by

Î ($456,400 2 $219,950)2 1 ($298,000 2 219,950)2 1 Á 1 ($108,000 2 219,950)2

12 2 1

5 $95,100
196 Chapter 5 Visualizing Variability

The sample standard deviation can be calculated in Excel using the STDEV.S function. In
Figure 5.20, the formula 5STDEV.S(A2:A13) in cell D8 calculates the standard deviation
of home sale values to be $95,100.
Standard deviation is a reliable measure of variability when the values of a variable
resemble the histogram in Figure 5.21, in which values are distributed symmetrically
around a single mode. For such bell-shaped distributions, we can use the standard deviation
to describe the variability of the distribution using intervals. Specifically,
●● < 68% of data values lie in the interval [mean 2 st. dev., mean 1 st. dev.]
●● < 95% of data values lie in the interval [mean 2 2 3 st. dev., mean 1 2 3 st. dev.]
●● . 99% of data values lie in the interval [mean 2 3 3 st. dev., mean 1 3 3 st. dev.]
However, because its calculation relies on the mean, the standard deviation can also
be heavily influenced by extreme values. For skewed distributions, the standard deviation
cannot be reliably used to provide an interpretable measure of the variability of a set of
values.
Another way to describe the variability of a set of values is with percentiles. A
There are many different ways percentile is the value of a variable which a specified (approximate) percentage of
to calculate the percentile observations are below that value. The pth percentile tells us the point in the data
for a data set. The method
described here matches the
where approximately p% of the observations have values less than the pth percentile;
method used by Excel, but hence, approximately (100 2 p)% of the observations have values greater than the pth
other software packages may percentile.
have slightly different ways of A percentile can be computed for any value between 0% and 100%, but common per-
calculating percentiles. centiles are the 25th, 50th, and 75th percentiles, which are also known as the first quartile,
second quartile, and third quartile, respectively. The 25th, 50th, and 75th percentiles are
called quartiles because they divide the data into four parts or quarters. The difference
between the third and first quartiles (the 75th and 25th percentiles) is often referred to as
the interquartile range, or IQR. The interquartile range spans the middle 50% of the dis-
tribution of a variable’s values and is sometimes used as a measure of variation.
To calculate the value of the pth percentile for a data set, we first compute its position
among the set of ordered values and then perform any necessary interpolation. To demon-
strate, consider the 25th percentile for the 12 values in the Cincinnati home sales data. The
position of the 25th percentile is computed by
25
3 (12 1 1) 5 3.25
100
The position of 3.25 for the 25th percentile means that it lies between 25% of the way between
the value of the third smallest and fourth smallest values. The third smallest value is $138,000
and the fourth smallest value is $142,000, so we compute the value of the 25th percentile as
$138,000 1 (3.25 2 3) 3 ($142,000 2 $138,000) 5 $139,000
Similarly, for the 50th percentile, the position is
50
3 (12 1 1) 5 6.5
100
The position of 6.5 for the 50th percentile means that it lies between 50% of the way
between the value of the sixth smallest and seventh smallest values. The sixth smallest
value is $199,500 and the seventh smallest value is $208,000, so we compute the value of
the 50th percentile as
$199,500 1 (6.5 2 6) 3 ($208,000 2 $199,500) 5 $203,750
Notice that the 50th percentile and the median have the same value. That is, 50% of
the observations have values less than the median, which matches its definition of the
median.
 5-2 Statistical Analysis of Distributions of Quantitative Variables 197

For the 75th percentile, the position is

75
3 (12 1 1) 5 9.75
100

The position of 9.75 for the 75th percentile means that it lies between 75% of the way
between the value of the ninth smallest and tenth smallest values. The ninth smallest value
is $254,00 and the tenth smallest value is $257,500, so we compute the value of the 50th
percentile as

$254,000 1 (9.75 2 9) 3 ($257,500 2 $254,000) 5 $256,625

The pth percentile can be calculated in Excel using the function PERCENTILE.EXC.
In Figure 5.20, the formula 5PERCENTILE.EXC(A2:A13, 0.25) in cell D10 calculates
the 25th percentile of home sale values to be $139,000. Similarly, the 50th percentile and
75th percentile are computed with the formulas =PERCENTILE.EXC(A2:A13, 0.5) and
5PERCENTILE.EXC(A2:A13, 0.75) in cells D11 and D12, respectively. Finally, the inter-
quartile range (IQR) is computed in cell D14 with the formula 5D122D10 to obtain a
value of $117,625.
The use of percentiles and the interquartile range to measure variability has advantages
over the range and standard deviation. First, extreme values do not distort the value of
percentiles. Second, percentiles do not require a variable’s distribution to be bell-shaped to
accurately convey its variability.

FIGURE 5.21 Bell-Shaped Distribution (Symmetric, Single Mode)

Box and Whisker Charts

Box and whisker charts are A box and whisker chart is a graphical summary of the distribution of data. A box and
also known as box plots. whisker chart is developed from the quartiles (the 25th, 50th, and 75th percentiles) of a set
of values. The following step-by-step directions illustrate how to create a box and whisker
chart in Excel using the file CincySales.
Step 1. Select cells A1:A13
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Statistic Chart button in the Charts group
When the list of statistic charts appears, select Box and Whisker
Further editing will result in a box and whisker chart similar to Figure 5.22. For explana-
tory purposes, we have added labels for the parts of the box and whisker chart, but these
are not necessary if the audience is familiar with the chart. As Figure 5.22 illustrates, a box
is drawn with the vertical ends of the box located at the first and third quartiles. For this
198 Chapter 5 Visualizing Variability

data, the first quartile is $139,000 and the third quartile is $256,625. This box contains the
middle 50% of the data. A horizontal line is drawn in the box at the location of the median
($203,750). An “X” marks the location of the mean ($219,950).
Vertical lines, called whiskers, extend from the top and bottom sides of the box. The
top whisker is drawn up to the largest value in the data that is less than or equal to third
quartile 11.5 3 IQR. For this data, the top whisker extends to $298,000, which is the larg-
est value less than or equal to $433,062.5 (5 $256,625 1 1.5 3 $117,625).
The bottom whisker is drawn down to the smallest value in the data that is greater than
or equal first quartile 21.5 3 IQR. For this data, the bottom whisker extends to $108,000,
which is the smallest value greater than or equal to 2$37,437.5 (5 $139,000 2 1.5
There is no single accepted 3 $117,625).
definition for what constitutes
A value outside the range [first quartile 2 1.5 3 IQR; third quartile 1 1.5 3 IQR]
an outlier in a data set.
Therefore, different software
is considered an outlier. For this data, only one value ($456,400) lies outside the range
packages may define outliers [2$37,437.5; $433,062.5]. This value is plotted in the box and whisker chart in Figure 5.22
slightly differently. to identify it as an outlier.

FIGURE 5.22 Box and Whisker Chart for Cincinnati Home Sales Data

Home Selling Prices ($)

500,000

450,000 Outlier

400,000

350,000

300,000
Whisker
250,000 3rd Quartile
Mean
200,000 IQR Median

150,000
1st Quartile
100,000

50,000

With their use of statistical measures, box and whisker charts support the detailed com-
parison of the distributions of multiple variables. We will use the SalesComparison file to
demonstrate the construction of box and whisker charts for multiple variables in Figure 5.23.
Step 1. Select cells B1:F11
Step 2. Click the Insert tab on the Ribbon
Step 3. Click the Insert Statistic Chart button in the Charts group
When the list of statistic charts appears, select Box and Whisker
Further editing will result in the visualization in Figure 5.23.
From the box and whisker charts in Figure 5.23, we can make several observations about
the distributions of home selling price in these five locations. Shadyside has the highest
home selling prices—the middle 50% of its price distribution is larger than all the selling
prices in Groton and Hamilton (and larger than almost every selling price in Fairview).
The median home selling price is nearly the same in Groton and Irving. However, Irving’s
houses have high variability in selling price, while Groton’s houses have low variability in
 5-2 Statistical Analysis of Distributions of Quantitative Variables 199

FIGURE 5.23 Box and Whisker Charts Comparing Multiple Variables

selling price. Irving’s selling price distribution is skewed right (positively skewed) by large
selling prices that extend to the largest over all five locations. Groton’s houses have relatively
similar selling prices, but there is one selling price in Groton that is an outlier due to its
relatively large value. The selling prices in Hamilton are generally the smallest over all five
locations. Hamilton’s selling price distribution demonstrates relatively little variability and is
nearly symmetric around its mean and median (which are the smallest of the five locations).

N otes 1 C omments

1. In our discussion of statistical measures of a variable’s distri- 4. For a set of n values ordered from smallest to largest,
bution, we have implicitly assumed that we have a sample x(1), x(2), Á x(n), the formula for computing the location, Lp,
of values that is a subset of the population of values. In of the pth percentile is
almost every application, it is not possible (or necessary) to p
collect data on the entire set of values of a variable. Lp 5 3 (n 1 1)
100
2. For a set of n values, x1, x2, Á xn, the formula for computing
Let :Lp ; be the largest integer less than or equal to Lp. Let
the sample mean is
LLp M be the smallest integer greater than or equal to Lp. Let
x1 1 x2 1 Á 1 xn x( :Lp; ) be the variable value at position :Lp ; when variables
x5
n are ordered from smallest to largest. Let x ( LLpM ) be the vari-
3. For a set of n values, x1, x2, Á xn, the formula for computing able value at position LLp M when variables are ordered from
the sample standard deviation is smallest to largest. Then, the value of the pth percentile is

Î
then given by
(x1 2 x )2 1 (x2 2 x )2 1 Á 1 (xn 2 x )2
s5
n21 pth percentile 5 x(:L ;) 1 (Lp 2 :Lp;) 3 (x(<L =) 2 x(:L ;))
p p p
200 Chapter 5 Visualizing Variability

5. A violin chart is an advanced visualization that combines provides a clearer picture of the distribution shape than
the statistical descriptors of a box and whisker chart with the box and whisker chart. As an example, the violin chart
a rotated and mirrored kernel density chart. Through its for the data in the DeathTwo file is as follows.
vertically displayed kernel density chart, the violin chart

Age At Death
125

100

75

50

25

Female Male

5-3 Uncertainty in Sample Statistics

In the first two sections of this chapter, we presented methods for visualizing the distribu-
tion of values for one or more variables. In this section, we discuss the visualization of the
variability that results from statistical sampling. The process of using sample data to make
estimates of or draw conclusions about one or more characteristics of a population is called
statistical inference.
A common example of statistical inference is political polling. Consider an example
in which members of a political party in Texas are considering the support of a particular
candidate for election to the U.S. Senate, and party leaders want to estimate the proportion
of registered voters in the state that favor the candidate. Suppose a sample of 400 registered
voters in Texas is selected, and 160 of those voters indicate a preference for the candidate.
Thus, an estimate of proportion of the population of registered voters who favor the candi-
date is 160/400 5 0.40. However, because this sample of 400 is only a portion of the voter
population of Texas, some error or deviation is to be expected between the sample propor-
tion and the population proportion that we are estimating. That is, there is uncertainty in
how close the sample proportion is to the population proportion.
Another example of statistical inference arises in market research. Consider the situa-
tion where a sample of weekly grocery bills is collected to estimate the average amount
of money spent on groceries by a target population of potential customers of a grocery
delivery service. Suppose a sample of 100 weekly grocery bills is selected, and the sample
mean is $102.70. However, because this sample of 100 is only a portion of possible weekly
grocery bills by potential customers, some error or deviation between the sample mean and
the population mean that we are estimating is to be expected. That is, there is uncertainty in
how close the sample mean is to the population mean.
In this section, we will discuss how to portray the uncertainty in sample-based estimates
of proportions and means. Failure to appropriately convey the inherent uncertainty in these
estimates may lead the audience to develop a false sense of confidence in these point esti-
mates. We want to construct visualizations that help the audience comprehend these sam-
ple-based estimates as intervals rather than points.
 5-3 Uncertainty in Sample Statistics 201

Displaying a Confidence Interval on a Mean

Because the sample mean cannot be expected to provide the exact value of the population
mean, a confidence interval is computed by adding and subtracting a value, called the
margin of error, to the sample mean:

sample mean 6 margin of error

The purpose of a confidence interval is to provide information about how close the sam-
ple mean may be to the value of the population mean. While the derivation of the formula
for the margin of error for a confidence interval on a mean is beyond the scope of this
book, we note that it is dependent on three factors: (1) the sample size, (2) how variable
the sample values are (as measured by the sample standard deviation), and (3) with how
much confidence we want to claim that the population mean lies within the interval. As the
sample size increases, the margin of error decreases. This is intuitive because as we collect
more data, we should be able to better estimate the mean. As the sample standard deviation
increases, the margin of error increases. This is intuitive because as the values in the data
demonstrate more variation, it becomes more difficult to estimate the mean. Finally, as
the required confidence level increases, the margin of error increases. If we must state an
interval with more confidence, then we must be more conservative with that interval and
increase its width. Common levels of confidence are 95% and 99%.
We provide an overview of Using the file DeathAvgAgeChart, we demonstrate the calculations of the margin of
the formula for computing error for a 95% confidence interval on a mean. In Figure 5.24, we compute the sample
the margin of error for a 95%
confidence interval on a mean
mean, the sample standard deviation, and sample size for the female and male observations
in the Notes+Comments at in the cell range E2:F4 using the appropriate Excel functions. In cells E5 and F5, we com-
the end of this section. pute the margin of error on the mean for the female and male samples, respectively, using
the CONFIDENCE.T function. The CONFIDENCE.T formula requires three input argu-
ments, 5CONFIDENCE.T(significance level, std dev, sample size). For a 95% confidence
interval, the first argument of the CONFIDENCE.T function is 1 2 0.95 5 0.05, and is
called the level of significance. The second and third arguments of the CONFIDENCE.T
function are the sample standard deviation and the sample size, respectively.
The 95% confidence interval on the mean age at death for females is 76.56 6 1.85 5
[74.71, 78.41]. We are 95% confident that the overall female population’s mean age at
death is in this interval. That is, if we collected 100 different samples of 327 females and
constructed confidence intervals on each of these 100 samples, we would expect 95 of the
100 confidence intervals to contain the overall female population’s mean age at death.
Similarly, the 95% confidence interval on the mean age at death for males is 70.85 6
1.81 5 [69.04, 72.66]. We are 95% confident that the overall male population’s mean age
at death is in this interval. That is, if we collected 100 different samples of 373 males and
constructed confidence intervals on each of these 100 samples, we would expect 95 of
the 100 confidence intervals to contain the overall male population’s mean age at death.
With the value of the margin of error computed, we can use Excel to display the confidence
intervals around a sample mean. Starting with the column chart in the file DeathAvgAgeChart,
the following steps show how to visualize a confidence interval for a mean.
Step 1. Click the column chart
Step 2. Click the Chart Elements button and select Error Bars
Click the black triangle to the right of Error Bars and select More Options...
DeathAvgAgeChart
Step 3. When the Format Error Bars task pane appears:
Click Error Bar Options
In the Error Amount area, select Custom
Click the Specify Value button to the right of Custom
In the Custom Error Bars dialog box, enter 5Data!$E$5:$F$5 in both the
Positive Error Value box and Negative Error Value box
Figure 5.25a displays the column chart for the mean age of death for females and males
with the vertical axis starting at zero. In most cases, starting the vertical axis at zero is
202 Chapter 5 Visualizing Variability

FIGURE 5.24 Calculations for Confidence Interval on Mean Using the File
DeathAvgAgeChart

FIGURE 5.25 Column Charts with Error Bars for Mean Age at Death:
Panel (a) Vertical Axis Starting at 0, Panel (b) Vertical Axis
Starting at 64

Females Live Longer on Average

Average Age at Death (years)
80 95% Confidence Interval

The legends labeling the error

70
bars in Figure 5.25 as a 95% 60
confidence intervals were
inserted manually as a text 50
boxes. 40
30
20
10
0
Female Male
(a)

Females Live Longer on Average

Average Age at Death (years)
80
78
By right-clicking the vertical 95% Confidence Interval
76
axis and selecting Format
Axis and then Axis Options, 74
we edited the minimum value
72
of the bound on the vertical
axis. 70
68
66
64
Female Male
(b)
 5-3 Uncertainty in Sample Statistics 203

recommended as this prevents distortion and misleading the audience. However, as the
objective of this column chart is to compare the difference (and not the absolute value of
the mean ages), starting the vertical axis at a non-zero value as in Figure 5.25b facilitates
this comparison. As Figure 5.25b clearly illustrates, the sample-based estimate of the
average age at death for females and males exhibits some uncertainty, but because the com-
puted confidence intervals do not overlap, based on this sample one can claim with at least
95% confidence that, on average, females live longer than males.

Displaying a Confidence Interval on a Proportion

Similar to the process of constructing a confidence interval around the sample mean, the degree
of uncertainty in the sample proportion can be expressed with a confidence interval formed by
adding and subtracting a margin of error. This confidence interval can be generally expressed as:

sample proportion 6 margin of error

The purpose of a confidence interval is to provide information about how close the
sample proportion may be to the value of the population proportion. The calculation of the
margin of error for a confidence interval on a proportion is different than the analogous
calculation for a confidence interval on a mean. While the derivation of the formula for the
margin of error is beyond the scope of this book, we note that it can be estimated using the
sample size, the sample proportion, and the confidence with which we want to claim that
the population proportion lies within the interval. A common level of confidence is 95%.
Using the file Incumbent, we demonstrate the calculations of the margin of error for a
95% confidence interval on a proportion (as displayed in Figure 5.26). The file Incumbent
We provide an overview of contains the yes or no response for 900 surveyed citizens on whether or not they sup-
the formula for computing port the incumbent president. In cell D2, we compute the sample size using the function
the margin of error for a
COUNTA to count all the text responses in the range A2:A901. In cell D3, we compute
95% confidence interval
on a proportion in the
the sample proportion as the ratio of “Yes” responses (counted with the Excel formula
Notes+Comments at the end 5COUNTIF(A2:A901, “Yes”)) and the sample size. Cell D4 contains the formula for the
of this section. margin of error for a 95% confidence interval on a proportion.
The 95% confidence interval on proportion of citizens who support the incumbent presi-
dent is 0.440 6 0.032 5 [0.408, 0.472]. We are 95% confident that the overall citizen pop-
ulation proportion of incumbent supporters is in this interval. That is, if we collected
100 different samples of 900 citizens and constructed confidence intervals on each of these
100 samples, we would expect 95 of the 100 confidence intervals to contain the overall
population’s proportion of incumbent supporters.

FIGURE 5.26 Calculations for Confidence Interval on Proportion Using

the File Incumbent
204 Chapter 5 Visualizing Variability

To create a visualization that accounts for the margin of error in the sample of 900 to
illustrate whether we are 95% confident that less than 50% of citizens support the incum-
bent president, we will employ the calculations in Figure 5.26 within a combo chart. First,
we need to arrange the source data for our combo chart. To create aesthetic visual spacing
in our chart, we will create three data series () in three different columns. As Figure 5.27
shows, we arrange the source data as three data series (Sample Proportion, Benchmark,
and Margin of Error) in three different columns, each with three entries. We arrange the
source data in cells C7:E9 in this manner as it will allow us to create an aesthetic visual
spacing in our final chart. In cells C7:C9, the first and third entries are “dummy” values of
zero and the second entry is the sample proportion. In cells D7:D9, all three entries cor-
respond to the benchmark value of 0.50 or 50% to which we want to compare the sample
proportion. In cells E7:E9, the first and third entries are “dummy” values of zero, and the
second entry is the margin of error for the confidence interval on the proportion.

FIGURE 5.27 Arranging Source Data for Chart Comparing Sample

Proportion to a Benchmark

With the source data in place, the following steps will show how to visualize a confi-
dence interval for a proportion and compare it to a 50% benchmark.
Step 1. Select cells C6:D9
Step 2. Click the Insert tab on the Ribbon
Incumbent
Step 3. Click the Insert Combo Chart button in the Charts group
When the list of combo charts appears, select Clustered Column - Line
Step 4. Click the chart and then click the Chart Elements button and select
Error Bars
Click the black triangle to the right of Error Bars and select More Options
Step 5. When the Add Error Bars dialog box appears:
Select Sample Proportion in the Add Error Bars based on Series: box
Click OK
Step 6 creates the error bar for Step 6. In the Format Error Bars task pane:
the column chart by extending Click Error Bar Options
the error bar line to a positive
In the Error Amount area, select Custom
deviation of 0.032 and to a
negative deviation of 0.032
Click the Specify Value button to the right of Custom
from the sample proportion In the Custom Error Bars dialog box, enter 5Data!$E$7:$E$9 in both
of 0.440. the Positive Error Value box and Negative Error Value box
Further editing will result in a histogram similar to Figure 5.28. As Figure 5.28 shows the
95% confidence interval on the proportion does not contain 0.50. Therefore, even account-
ing for the margin of error in the sample proportion, we are 95% confident that less than
50% of citizens support the incumbent president.
 5-4 Uncertainty in Predictive Models 205

FIGURE 5.28 Combo Chart with Error Bars on Sample Proportion for
Incumbent Data
President Has Less Than 50% Support
Proportion Supporting 95% Confidence Interval

0.6

0.5

0.4

0.3

0.2

0.1

0.0

N otes 1 C omments

1. For a set of n values where x is the sample mean and s is 3. Determining whether a statistically significant difference
the sample standard deviation, the formula for computing between two means exists by checking for overlap is not
a 95% confidence interval on the mean is approximately: as precise as computing a single confidence interval on
s the difference between the means. However, comparing
x 6 1.96
Ïn the confidence intervals on the respective means is visually
2. For a set of n values where p is the sample proportion, the appealing and will not result in stating a claim that is not
formula for computing a 95% confidence interval on the true at the specified level of confidence.
proportion is approximately:
Ïp(1 2 p)
p 6 1.96
Ïn

5-4 Uncertainty in Predictive Models

Predictive analytics consists of techniques that use models constructed from past data to
predict the value of future observations. For example, past data on product sales may be
used to construct a mathematical model to predict future sales. This model can factor in the
product’s growth trajectory and seasonality based on past patterns.
In this section, we consider predictive models that provide point estimates of future observa-
tions. Similar to the sample-based estimates of statistical inference discussed in Section 5.3, it
is unrealistic to expect that these predicted point estimates have no error. That is, there is uncer-
tainty in how close a predicted point estimate will be to the corresponding future observation.
The uncertainty in a model’s predicted values of future observations can be expressed
using a prediction interval. A prediction interval for a future observation is conceptually
similar to a confidence interval on a population mean or proportion, but is computed accord-
ing to a different formula. In this section, we consider the visualization of prediction intervals
for two different types of predictive models: simple linear regression and time series models.

Illustrating Prediction Intervals for a Simple Linear Regression Model

Simple linear regression approximates the relationship between two variables with a straight
The dependent variable is also
line. The variable being predicted is called the dependent variable (y). The dependent variable is
sometimes referred to as the plotted on the vertical axis. The variable being used to predict or explain the dependent variable is
response or target variable. called the independent variable (x). The independent variable is plotted on the horizontal axis.
Yourier LLC is a home delivery service that picks up items from stores and delivers
them to customers. To assess the effectiveness of its process of routing customer requests,
206 Chapter 5 Visualizing Variability

Yourier is interested in predicting the travel time of a route based on the number of requests
on a route. Therefore, travel time is the dependent variable and the number of requests is
the independent variable. For a sample of 10 routes, Figure 5.29 displays a scatter chart
with number of requests on the horizontal axis (x) and travel time on the vertical axis (y).

FIGURE 5.29 Scatter Chart of Travel Time versus Number of Requests

Yourier LLC Route Performance

Travel Time (hours)
3.5

3.0

2.5

2.0

1.5

1.0

0.5

0
0 1 2 3 4 5 6 7
Number of Requests

Yourier’s predictive analytics team has constructed a simple regression model to predict
a route’s travel time based on the number of requests served by the route. Based on these
10 routes, the simple regression equation is

travel time 5 0.2852 3 number of requests 1 1.356

The calculation of prediction
intervals is beyond the scope
Figure 5.30 lists the data in YourierChart. Columns B and C contain the number of requests
of this book. Many dedicated
statistical software packages
and travel time for the routes, respectively, that are used to create the scatter chart in Figure 5.29.
will automatically provide this Column D lists the predictions using this simple regression equation for the routes. In addition,
output for prediction models. columns E and F list the lower and upper 95% prediction interval limits for the routes.

FIGURE 5.30 Simple Regression Predictions and 95% Prediction Interval

Limits for Yourier Data
 5-4 Uncertainty in Predictive Models 207

Using the data on the point predictions and the lower and upper limits of the 95% predic-
tion intervals, the following steps demonstrate how to visualize this prediction information
on the scatter chart of travel time versus number of requests.
Step 1. Right click the chart and select Select Data…
Step 2. When Select Data Source dialog box appears, click the Add button Add

Step 3. In the Edit Series dialog box:

YourierChart
Enter 5Data!$D$1 in the box under Series name:
Enter 5Data!$B$2:$B$13 in the box under Series X values:
Enter 5Data!$D$2:$D$13 in the box under Series Y values:
Click OK
Step 4. In the Select Data Source dialog box, click the Add button Add

Step 5. In the Edit Series dialog box:

Enter 5Data!$E$1 in the box under Series name:
Enter 5Data!$B$2:$B$13 in the box under Series X values:
Enter 5Data!$E$2:$E$13 in the box under Series Y values:
Click OK
Step 6. In the Select Data Source dialog box, click the Add button Add

Step 7. In the Edit Series dialog box:

Enter 5Data!$F$1 in the box under Series name:
Enter 5Data!$B$2:$B$13 in the box under Series X values:
Enter 5Data!$F$2:$F$13 in the box under Series Y values:
Click OK
Click OK to close Select Data Source dialog box
At this stage, the visualization consists of four series of data plotted as scatter charts as in Figure 5.31.

FIGURE 5.31 Scatter Chart Display of Observations, Predictions, and

Prediction Interval Limits

Yourier LLC Route Performance

Travel Time (hours)
5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0
0 1 2 3 4 5 6 7 8
Number of Requests

The next steps modify the visualization to display the point predictions and prediction
interval limits with lines rather than points.
Step 8. Click a data series corresponding to either the point predictions or the predic-
tion interval limits. With this data series selected, right-click, and then select
Change Series Chart Type… from the menu
Step 9. When the Change Chart Type dialog box appears, click X Y Scatter XY (Scatter)

Select Scatter with Smooth Lines on the All Charts tab

Click OK
208 Chapter 5 Visualizing Variability

Make sure that only the blue Step 10. Click the blue line representing the observed travel times. Right click this line
line is selected. and select Format Data Series…
Step 11. In the Format Data Series task pane, click the Fill & Line icon
Click Line and select No Line
Click Marker and under Marker Options select Automatic
Step 12. Select the chart and click the Chart Elements button and select Legend
Click the black triangle to the right of Legend and select Top
Further editing will result in a chart similar to Figure 5.32. We use color to differentiate
between the past observations (blue data points), point estimates from regression model
(solid orange line), and the limits of the prediction intervals (dashed orange lines).
The 95% prediction interval corresponds to the range that we are 95% confident will
contain the value of the independent variable in a future observation with a specified value
of the dependent variable. For instance, for a future route servicing three requests, the sim-
ple regression model predicts a travel time of 2.211 hours and is 95% confident that the
route’s travel time will be between 1.354 hours and 3.069 hours (a width of 1.715 hours).
Now consider a future route servicing six requests. The simple regression model predicts
a travel time of 3.067 hours and is 95% confident that the route’s travel time will be between
2.065 hours and 4.069 hours (a width of 2.004 hours). Intuitively, the simple regression model
predicts a route with more requests requires more travel time. In addition, notice that the
simple regression model is more confident in its travel time predictions for routes with three
requests than for routes with six requests. We can observe this in Figure 5.32 by noting that
the dotted lines corresponding to the prediction interval limits are not straight lines. Instead,
these lines are slightly curved to depict that the width of the prediction interval is the narrow-
est near the mean value of number of requests. That is, the width of the prediction interval
depends on the value of the independent variable for the observation being predicted.

FIGURE 5.32 Combo Chart Displaying 95% Prediction Intervals on Travel Time

Yourier LLC Route Performance

Travel Time (hours)
5.0
Upper 95% P.I.
4.5

4.0

3.5
Prediction
3.0

2.5

2.0 Lower 95% P.I.

1.5

1.0

0.5

0
0 1 2 3 4 5 6 7 8
Number of Requests

Illustrating Prediction Intervals for a Time Series Model

Time series data is a sequence of observations on a variable measured at successive points
in time. The measurements may be taken every hour, day, week, month, year, or at any
other regular interval. To display a time series chart, a special type of line chart called a
time series chart is typically used. In a time series chart, the time unit is represented on
 5-4 Uncertainty in Predictive Models 209

the horizontal axis and the values of the variable are shown on the vertical axis. Connecting
the consecutive observations with line segments in a time series chart accentuates the tem-
poral nature of the data and the inherent relationship between consecutive time periods.
Bundaberg Brewed Drinks is a company that makes craft brewed sodas. Premium
Australian root beer is an integral part of its product portfolio, and Bundaberg is interested in
forecasting its sales in future quarters. Figure 5.33 displays a time series chart of Bundaberg’s
sales of root beer measured over the past 17 quarters. From this time series chart, the sea-
sonal nature of root beer sales is evident. Sales peak in quarters 3, 7, 11, and 15, while sales
are lowest in quarters 1, 5, 9, 13, and 17. That is, the seasonal pattern for root beer sales
repeats itself every fourth quarter. In addition, the time series chart suggests that there is no
meaningful trend in root beer sales. That is, apart from seasonal variation, quarterly sales do
not appear to exhibit any upward or downward change. This can be seen by comparing the
sales from every fourth quarter and observing that there is no upward or downward pattern.

FIGURE 5.33 Time Series Chart of Quarterly Root Beer Sales

Bundaberg Root Beer Sales

Sales ($1000s)
250
200
150
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Quarter

The calculation of predictions

and prediction intervals from
time series models is beyond Bundaberg’s forecasting team has constructed a time series model that accounts for the
the scope of this book. Many
seasonality to predict root beer sales over the next eight quarters. Figure 5.34 lists the data
dedicated statistical software
packages will automatically
in BundabergChart. Column C contains the point estimates for root beer sales for the next
provide this output for eight quarters. Columns D and E list the lower and upper 95% prediction interval limits for
prediction models. these eight predictions.
Using the data on the point predictions (in column C) and the lower and upper limits of
the 95% prediction intervals (in columns D and E), the following steps demonstrate how to
visualize this prediction information on the time series chart of quarterly root beer sales.
Step 1. In cell C18, enter 5B18
Step 2. Right click the chart and select Select Data…
Step 3. When Select Data Source dialog box appears, click the Add button Add
BundabergChart
Step 4. In the Edit Series dialog box:
In step 1, we set the
enter =Data!$C$1 in the box under Series name:
prediction for quarter 17 to be enter 5Data!$C$2:$C$26 in the box under Series values:
the same as the actual sales in Click OK
order to create a connected Step 5. In the Select Data Source dialog box, click the Add button Add
line between the actual sales Step 6. In the Edit Series dialog box:
and the forecasted sales.
enter 5Data!$D$1 in the box under Series name:
enter 5Data!$D$2:$D$26 in the box under Series values:
click OK
Step 7. In the Select Data Source dialog box, click the Add button Add

Step 8. In the Edit Series dialog box:

enter 5Data!$E$1 in the box under Series name:
Click OK to close Edit Series dialog box
enter 5Data!$E$2:$E$26 in the box under Series values:
click OK
Click OK to close Select Data Source dialog box
210 Chapter 5 Visualizing Variability

FIGURE 5.34 Time Series Data and Predictions for Quarterly Root Beer
Sales

FIGURE 5.35 Line Chart Illustrating Predictions for Future Quarterly Root
Beer Sales

Bundaberg Root Beer Sales

Sales ($1000s)
Sales Prediction Lower 95% P.I. Upper 95% P.I.
250
200
150
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Quarter
Glossary 211

Step 9. Select the chart and click the Chart Elements button , select Legend
Click the black triangle to the right of Legend, and select Top
Further editing will result in a chart similar to Figure 5.35. We use color to differentiate
between the sales from past quarters (solid dark blue line), point estimates for future quar-
ters (solid light blue line), and the limits of the prediction intervals (dashed light blue lines).
As Figure 5.35 illustrates, the predictions for root beer sales in the next eight quarters
reflect the seasonal nature of this product. In general, a 95% prediction interval corre-
sponds to the range that we are 95% confident will contain the time series variable value
for the specified future time period. For instance, the time series model predicts root beer
sales of $119,000 in quarter 18 and is 95% confident that sales will be between $81,100
and $156,900 (a width of $75,800).
Now consider the forecast for quarter 22. As quarter 22 is in the same season as quar-
ter 18 and past sales do not exhibit any upward or downward trend, the time series model
again predicts root beer sales to be $119,000 (same as its prediction for quarter 18).
However, the 95% prediction interval on quarter 22 sales is $20,400 to $127,600 (a width
of $107,200), which is wider than the prediction interval for quarter 18. To maintain a
level of 95% confidence, the time series model must quote a wider prediction interval as it
makes predictions further into the future. That is, the width of the prediction interval for a
time series model depends on how far into the future the prediction is. This is reflected by
the growing distance between the lower and upper limits of the 95% prediction interval in
Figure 5.35 for predictions further in the future.

S U M M A RY

In this chapter, we have discussed how to visualize the variation in the values of a variable
of interest. We introduced the notion of a frequency distribution as measured by counts and
relative (percent) frequency. Then we showed how to use Excel to visualize a frequency
distribution for a categorical variable.
We introduced several ways to visualize a frequency distribution for a quantitative vari-
able. We demonstrated how to use the columnar display of histograms to analyze the shape
of a distribution and defined the measure of skewness to formally describe distribution
shape. As an alternative to a histogram, we explained how to use a line chart to create a
frequency polygon to illustrate the shape of variable’s distribution. For small data sets, we
introduced strip charts and how to use hollow dots and jittering to avoid occlusion.
We defined formal statistical measures of central location such as the mean, median,
and mode. Then we defined formal statistical measures of variability such as the range,
standard deviation, and interquartile range. We showed how to construct a box and whisker
chart and interpret the statistical measures utilized by this chart.
In the final two sections, we discuss how to convey uncertainty that arises in statis-
tical inference and predictive analytics. Specifically, we describe how to use error bars
to portray the margin of error in sample-based estimates of a mean or a proportion. For
predictions based on a simple linear regression model, we explain how to visualize pre-
diction intervals generated by this causal model. Finally, we consider time series data
and show how to display prediction intervals on forecasts from a time series model.

G L O S S A RY

Benford’s Law An observation that the leading digit in many naturally occurring data
sets approximately obeys a known frequency distribution. In particular, the leading digit is
likely to be small with 1 being the most likely leading digit and 9 the least likely.
Bins The nonoverlapping groupings of data used to create a frequency distribution. Bins
for categorical data are also known as classes.
Box and whisker chart A graphical summary of data based on the quartiles of a distribution.
Categorical variable Data for which categories of like items are identified by labels or
names. Arithmetic operations cannot be performed on categorical variables.
212 Chapter 5 Visualizing Variability

Confidence interval An estimate of a population parameter that provides an

interval of the form point estimate 6 margin of error, believed to contain the value
of the parameter with a specified confidence. For example, if an interval estimation
procedure provides intervals such that 95% of the intervals formed using the procedure
will include the population parameter, the interval estimate is said to be constructed at
the 95% confidence level.
Dependent variable The variable that is being predicted or explained. It is generally plotted
on the vertical axis. Also sometimes referred to as the response variable or target variable.
Frequency distribution A summary of data that shows the number (frequency) of
observations in each of several nonoverlapping bins (classes).
Frequency polygon A chart used to display a distribution by using lines to connect the
frequency values of each bin.
Histogram A columnar presentation of a frequency distribution, relative frequency
distribution, or percent frequency distribution of quantitative data constructed by placing
the bin intervals on the horizontal axis and the frequencies, relative frequencies, or percent
frequencies on the vertical axis.
Independent variable The variable used for predicting values of the dependent variable.
It is generally plotted on the horizontal axis.
Interquartile range The difference between the third and first quartiles.
Jittering The process of slightly altering the actual values of one or more variables in the
observations of a data set so that identical observations occupy slightly different positions
when plotted.
Kernel density chart A chart for visualizing the distribution that smooths the bin
frequency values of a histogram representation by using kernel density estimation.
Margin of error The value added to and subtracted from a point estimate in order to
develop a confidence interval for a population parameter.
Mean A measure of central location computed by summing the data values and dividing
by the number of observations.
Median A measure of central location provided by the value in the middle when the data
are arranged in ascending order. The median is the 50th percentile.
Mode A measure of central location defined as the value (or range of values) that occurs
with the greatest frequency.
Occlusion The inability to distinguish some individual data points because they are hidden
behind others with the same or nearly the same value.
Outlier An unusually small or unusually large data value.
Percent frequency A frequency measure in a distribution analysis that computes the
percentage of observations in each of several nonoverlapping bins (classes).
Percentile A value such that approximately p% of the observations have values less than
the pth percentile; hence, approximately (100 2 p)% of the observations have values
greater than the pth percentile.
Population The set of all elements of interest in a particular study.
Power law An observation that in some data the relative change in one variable results in a
proportional relative change in another variable.
Prediction interval An interval estimate of the prediction of a future value of the
dependent variable such that there is a specified confidence that this interval will contain
the future value of the dependent variable.
Predictive analytics Techniques that use models constructed from past data to predict the
future or to ascertain the impact of one variable on another.
Probability distribution A description of the range and relative likelihood of possible
values of a random variable.
Quantitative variable Data for which numerical values are used to indicate magnitude,
such as how many or how much. Arithmetic operations such as addition, subtraction, and
multiplication can be performed on a quantitative variable.
Quartile The 25th, 50th, and 75th percentiles, referred to as the first quartile, second
quartile, and third quartile, respectively. The quartiles can be used to divide a set of data
values into four parts, with each part containing approximately 25% of the values.
Problems 213

Random variable A quantity whose values are not known with certainty.
Range A measure of variability defined to be the largest value minus the smallest value.
Relative frequency A frequency measure in a distribution analysis that computes the
fraction or proportion of observations in each of several nonoverlapping bins (classes).
Sample A subset of the population.
Simple linear regression A statistical procedure predicting the value of one dependent
variable with the value of one independent variable through a linear equation.
Skewness A measure of the lack of symmetry in a distribution.
Standard deviation A measure of variability that captures how much a set of values
deviates from the mean.
Statistical inference The process of making estimates and drawing conclusions about
one or more characteristics of a population (the value of one or more parameters) through
analysis of sample data drawn from the population.
Strip chart A chart consisting of sorted variable values along either the horizontal or
vertical axis.
Time series chart A chart where a measure of time is represented on the horizontal axis
and a variable of interest is shown on the vertical axis. Temporally consecutive data points
are generally connected with straight lines.
Time series data Data that are collected at intervals over time.
Trellis display A vertical or horizontal arrangement of individual charts of the same type,
size, scale, and formatting that differ only by the data they display.
Variation Differences in values of a variable over observations.
Violin chart A graphical method that encases the elements of a box and whisker chart
inside a rotated and mirrored kernel density chart.

P R O B L E M S

Conceptual
1. Histogram Bin Widths for Boating Customers. Based on a survey of 1,046 indi-
viduals who recently took the boat cruise on Lake Havasu, Rochambeau Boating is
analyzing the percent frequency distribution of its customers’ ages. An analyst has
created four histograms by varying the bin size. Rochambeau seeks a visualization of
the customer age distribution that captures general trends in the data but does not blur
patterns by grouping customers with disparate ages (and therefore behaviors) into the
same bins. Which histogram would you recommend that the analyst use to describe the
customer age distribution? LO 2
i.
Age Distribution of Customers
Percent Frequency
5.0%
4.5%
4.0%
3.5%
3.0%
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81

Age (years)
214 Chapter 5 Visualizing Variability

ii.
Age Distribution of Customers
Percent Frequency
20%
18%
16%
14%
12%
10%
8%
6%
4%
2%
0%

(20, 25]
(25, 30]

(40, 45]

(45, 50]

(65, 70]
(10, 15]
(15, 20]

(50, 55]
(55, 60]
(60, 65]

(70, 75]
(75, 80]
[0, 5]
(5, 10]

(30, 35]
(35, 40]
Age (years)

iii.
Age Distribution of Customers
Percent Frequency
40%
35%
30%
25%
20%
15%
10%
5%
0%
[0, 10] (10, 20] (20, 30] (30, 40] (40, 50] (50, 60] (60, 70] (70, 80]
Age (years)

iv.
Age Distribution of Customers
Percent Frequency
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
[0, 15] (15, 30] (30, 45] (45, 60] (60, 75] (75, 90]
Age (years)
Problems 215

2. Stacked Column Chart for Boating Customers. Based on a survey of 1,046 individ-
uals who recently took the boat cruise on Lake Havasu, Rochambeau Boating is ana-
lyzing the demographics of its customers. It is believed that the sample (388 females
and 658 males) is representative of Rochambeau’s overall customer population. An
analyst has created the following chart depicting the age and sex distribution of the
survey respondents. LO 3
Age Distribution of Customers
Percent Frequency
40% Male Female
35%
30%
25%
20%
15%
10%
5%
0%

(25, 30]

(30, 35]

(35, 40]

(40, 45]

(45, 50]

(65, 70]
[0, 5]

(5, 10]

(10, 15]

(15, 20]

(20, 25]

(70, 75]

(75, 80]
(50, 55]

(55, 60]
(60, 65]
Age (years)

Which of the following statements accurately assess this stacked column chart? (Select
all that apply.)
i. The use of color is unnecessary and distracting.
ii. The stacked orientation makes it difficult to compare the shape of the age distri-
butions of male and female customers.
iii. A strength of this chart is how the stacked orientation visualizes the overall age
distribution and how the number of customers in each age bin are split into male
and female categories.
iv. It would be better to orient this chart vertically as a stacked bar chart.
3. Clustered Column Chart for Boating Customers. Based on a survey of
1,046 individuals who recently took the boat cruise on Lake Havasu, Rochambeau
Boating is analyzing the demographics of its customers. It is believed that the sample
(388 females and 658 males) is representative of Rochambeau’s overall customer
population. An analyst has created the following chart depicting the age and sex
distribution of the survey respondents. LO 2, 3
Age Distribution of Customers
Percent Frequency
25% Male Female

20%

15%

10%

5%

0%
[0, 5]

(5, 10]

(10, 15]

(15, 20]

(20, 25]

(25, 30]

(30, 35]

(35, 40]

(40, 45]

(45, 50]

(50, 55]

(55, 60]

(60, 65]

(65, 70]

(70, 75]

(75, 80]

Age (years)
216 Chapter 5 Visualizing Variability

a. Which of the following statements accurately criticizes this clustered column chart?
i. There are too few bins.
ii. The alternating male and female frequency columns cast a disrupted visualization
of the distributions.
iii. The use of color is unnecessary and distracting.
iv. A stacked column chart would compare females and males better.
b. Which of the following is a better way to visualize these data to facilitate the com-
parison of the age distributions of female and male customers?
i. Stacked column chart
ii. Strip chart
iii. Frequency polygon
iv. Stem and leaf chart
4. Age Pyramid for Boating Customers. Based on a survey of 1,046 individuals who
recently took the boat cruise on Lake Havasu, Rochambeau Boating is analyzing the demo-
graphics of its customers. Treating the sample of 388 females and 658 males separately,
an analyst is interested in comparing how the ages of the female customers are distributed
as a percentage of the 388 females surveyed, and how the ages of the male customers are
distributed as a percentage of the 658 males surveyed. An analyst has created the following
chart depicting the age and sex distribution of the survey respondents. LO 3

Age Distribution of Customers

Age (years)
(75, 80] Female Male
(70, 75]
(65, 70]
(60, 65]
(55, 60]
(50, 55]
(45, 50]
(40, 45]
(35, 40]
(30, 35]
(25, 30]
(20, 25]
(15, 20]
(10, 15]
(5, 10]
[0, 5]
20% 15% 10% 5% 0% 5% 10% 15% 20% 25%
Percent Frequency

a. Which of the following statements accurately criticizes this chart?

i. Orienting the frequency bars for females and males end-to-end makes it harder to
visually compare their magnitudes.
ii. There are too many bins.
iii. The use of color is unnecessary and distracting.
iv. Rotating the chart by 90 degrees would allow for better binwise comparison of
females and males.
b. Which of the following is a better way to visualize these data to facilitate the com-
parison of the age distributions of female and male customers?
i. Clustered column chart
ii. Strip chart
iii. Violin chart
iv. Frequency polygon
5. Distribution Deviation Chart for Boating Customers. Based on a survey of 1,046
individuals who recently took the boat cruise on Lake Havasu, Rochambeau Boating is
analyzing the demographics of its customers. Treating the sample of 388 females and
658 males separately, an analyst is interested in comparing how the ages of the female
customers are distributed as a percentage of the 388 females surveyed, and how the
ages of the male customers are distributed as a percentage of the 658 males surveyed.
Problems 217

An analyst has created the following chart computing the difference in these percent-
ages for each age bin. LO 3

Percent of Females - Percent of Males

Difference in Percentages
4%

3%

2%

1%

0%

–1%

–2%

–3% [0, 5]

(5, 10]

(10, 15]

(15, 20]

(20, 25]

(25, 30]

(30, 35]

(35, 40]

(40, 45]

(45, 50]

(50, 55]

(55, 60]

(60, 65]

(65, 70]

(70, 75]

(75, 80]
Age (years)

a. Which of the following statements accurately criticizes this chart?

i. There are too few bins.
ii. There are too many bins.
iii. The chart needs to use color differentiation.
iv. This chart provides no insight on the shape of the age distributions for female and
male customers, and the differences computed are highly dependent on the bins used.
b. Which of the following is a better way to visualize these data to facilitate the com-
parison of the age distributions of female and male customers?
i. Kernel density chart
ii. Clustered column chart
iii. Frequency polygon
iv. Strip chart
6. Daily Temperature Strip Charts. Aspiring meteorologist Jesse Kosch has collected
data on the daily high temperature in Valentine, Nebraska, over an entire year. Jesse
then used a strip chart to display each day’s high temperature by month. LO 4
Temperature (degrees Fahrenheit)

75

50

25

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
218 Chapter 5 Visualizing Variability

a. Which of the following statements accurately criticizes this chart?

i. It is not clear what each data point represents.
ii. The chart suffers from occlusion as there is not a distinct data point for each day
in each month.
iii. A line chart would better display this data.
iv. The chart needs to use color differentiation.
b. How could Jesse best improve this visualization?
i. Differentiate each month with a different color
ii. Jitter each data point vertically
iii. Change horizontal axis to day of year and plot as time series chart
iv. Use hollow dots instead of solid dots and horizontally jitter each data point
7. Histogram for Assessed Home Values. Chris Fultz, a realtor at Betelgeuse Realty, is
conducting research on assessed home values in a local suburb. Chris has constructed
the following histogram. LO 2, 5, 6
Distribution of Assessed Home Values
Frequency Count
140
120
100
80
60
40
20
0
63 ]
17 ]
71 ]
25 ]
79 ]
33 ]
87 ]
41 ]
95 ]
49 ]
03 ]
57 ]
11 ]
65 ]

]
09 ]
(1 163
(2 217
(2 271
(3 325
(3 379
(4 433
(4 487
(5 541
(5 595
(6 649
(7 703
(7 757
(8 811
(8 865
19
(1 109

,9
,
,
,
,
,
,
,
,
,
,
,
,
,
,
5,
[5

Home Value ($1000s)

Which box and whisker chart corresponds to the histogram?

i.

ii.
Problems 219

iii.

iv.

8. Frequency Polygon for Assessed Home Values. Abbie Aburizek, a realtor at Alamo
Acres, is conducting research on assessed home values in a local suburb. Abbie has
constructed the following frequency polygon. LO 2, 5, 6

Distribution of Assessed Home Values

Frequency Count
100
90
80
70
60
50
40
30
20
10
0
]

13 ]

30 ]
(3 47]

64 ]
(3 81]

]
(2 96]
]
45 5]
]

62 ]
28

13

30

64

98
79
11

62
4
,2

,3

,3

,3

,3

,3

,3
,2
,2
,2
,2

,2
11

96

47

81
79
28
94
(2

(3

(3

(3
(2
(2
(2
[1

(2

Home Value ($1000s)

220 Chapter 5 Visualizing Variability

Which box and whisker chart corresponds to the frequency polygon?

i.

ii.

iii.

iv.
Problems 221

9. Life Insurance Analysis. Josh Bell, an actuarial scientist for Yolo Life Insurance, has
created the following box and whisker charts of the age at death for a randomly drawn
sample representative of potential clients. LO 6

Which of the following assessments are true?

i. The distributions are positively skewed; females live longer on average.
ii. The distributions are negatively skewed with several outliers of extremely small
values; females live longer on average.
iii. The distributions are symmetric; the mean and median values are the same.
iv. There are no outliers in either sample; both the female and male distributions
have the same shape.
10. Restaurant Delivery Times. A foodie blog has gathered data on the delivery times for
four various restaurant delivery services (CHOMP, Uber Eats, DoorDash, Grubhub).
Blog editor Auguste Gusteau has been analyzing the distribution of delivery times for
each of these delivery companies and is contemplating how to visually compare the
four delivery time distributions. Which of the following would you recommend? LO 3
i. Plotting all four delivery time distributions on a clustered column chart.
ii. Using a trellis display arranging the four frequency polygons vertically.
iii. Using a collection of six charts each displaying a different pair of delivery time
distributions with frequency polygons.
iv. Plotting all four delivery time distributions with frequency polygons.
11. Call Center Response Times. Skylar Diggins is reviewing the performance of four
call centers under her management. Twelve observations of the amount of time it took
to process refund requests have been collected from each of the four call centers. Based
on these observations, the following column chart displays the average service time of
each call center. LO 7
Average Time to Process Refund Requests
Processing Time (minutes)
35
30

25

20

15

10

0
A B C D
Call Center
222 Chapter 5 Visualizing Variability

a. Why may the interpretation of the average service time from this column chart be misleading?
b. How can this chart be improved to provide a more accurate comparison of the aver-
age service times at the call centers?
12. Vision Insurance. Before including vision insurance in the benefits package for its
10,000 employees, Valmont Industries wants to confirm that it is desired by a major-
ity of employees. Valmont has taken an employee survey and out of the 100 survey
respondents, 55 employees said they would opt into the coverage. An analyst has dis-
played the results in the following column chart. LO 7

Proportion of Employees Wanting Vision Insurance

0.6

0.5

0.4

0.3

0.2

0.1

a. Why may the results of the sample proportion in this chart be misleading in terms
of interpreting the proportion of all 10,000 employees who want vision insurance?
b. How can this chart be changed to convey the survey results more accurately?
13. Restaurant Franchising. A restaurant chain particularly popular with college students has
collected data on the relationship between quarterly sales of existing restaurants and the
size of the college student population in the respective restaurant’s immediate region. Janice
Moore, director of franchising, has constructed a simple linear regression model using quar-
terly sales as the dependent variable (y) and college student population as the independent
variable (x). The data and simple linear regression model are displayed in the chart below.

Performance of Restaurant Locations

Quarterly Sales ($1000s)
225

y = 5x + 60
150

110

75

0
0 5 10 15 20 25 30
Student Population (1000s)
 Problems 223

The development team has identified a promising location for a new franchise that has
10,000 college students living nearby. Based on her simple regression model, Janice
predicts that this new franchise would experience quarterly sales of Y 5 5x 1 60 5
(5 3 10) 1 60 5 110, or $110,000. Janice has highlighted this prediction on her chart.
LO 8
a. Why may the results of this chart be misleading?
b. How can this chart be changed to convey the survey results more accurately?
14. Forecasting Stock Price. To inform his day trading, Jorge Belfort has gathered his-
torical stock price data on Nile Inc., a green energy company he has been researching.
Jorge has constructed the following time series chart to display the stock prices as well
as his forecasts of the stock price over the next few days.
Nile Inc.
Stock Price ($ per share) Price Forecast

200

150

100

50

0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57
Day

How can Jorge modify this chart to further communicate any uncertainty associated
with his forecast for this stock? LO 8
Applications

15. Most Visited Websites. In a recent report, the top five most-visited English-language
websites were google.com (GOOG), facebook.com (FB), youtube.com
Websites (YT), yahoo.com (YAH), and wikipedia.com (WIKI). The file WebSites contains a
sample of the favorite website for 50 Internet users. LO 1
a. Using a column chart, visualize the frequency distribution for these data.
b. On the basis of the sample, which website is listed most frequently as the favorite
website for Internet users? Which is second?
16. Auditing Travel Expense Reports. Courtney Boyce is currently auditing a sample of
the travel expense reports submitted by company employees over the past year. LO 1
a. Using the data in the file TravelExpenses, create a percent frequency distribution of
TravelExpenses
the first digit of the expense reported in each report.
b. Assessing this distribution with Benford’s Law, does Courtney have any reason to
suspect reporting error or fraud? Hint: Treat the first digit of expense amount as a
categorical variable (text) rather than as a numerical value.
17. University Endowments. University endowments are financial assets that are donated
by supporters to be used to provide income to universities. There is a large discrepancy
in the size of university endowments. The file Endowments provides a listing of many
Endowments
of the universities that have the largest endowments as reported by the National Associ-
ation of College and University Business Officers. LO 2, 5
a. Construct a histogram of the frequency counts.
b. Comment on the shape of the distribution displayed by the histogram.
18. Busiest North American Airports. Based on the total passenger traffic, the airports in
the file Airports are among the busiest in North America. LO 2, 5
a. Construct a histogram of the frequency counts using a bin width of 10 with the first
Airports
bin starting at a value of 30 million.
b. What is the most common passenger traffic range based on this histogram?
c. Describe the shape of the histogram.
224 Chapter 5 Visualizing Variability

19. Television Viewership of Football Games. Companies considering purchasing ads

during televised XFL football games are interested in the viewership of games as a way
of estimating the exposure their ads get. The file PrimeTimeXFL contains a sample of
PrimeTimeXFL
viewership numbers for XFL games televised on Friday and Saturday nights, respec-
tively. L0 6
a. Construct box and whisker charts on the same chart that compares the viewership
distribution of Friday night XFL games to the viewership distribution of Saturday
night XFL games.
b. Use the box and whisker charts to compare the two distributions.
c. Use the box and whisker chart of the Friday night XFL games to describe this
distribution.
d. Use the box and whisker chart of the Saturday night XFL games to describe this
distribution.
20. CEO Time in Meetings. A study was conducted to examine how chief executive officers
(CEOs) spend their work days. The file CEOtime contains a sample of data from this
study, which tracked the minutes per day various CEOs spent in meetings. LO 4, 5
CEOtime
a. Construct a strip chart displaying the data, making sure to employ techniques to
avoid occlusion.
b. What is the median amount of time a CEO spends in meetings per day?
c. What is the mean amount of time a CEO spends in meetings per day?
d. What is the most common amount of time a CEO spends in meetings per day?
21. Number of Hours Worked per Week by Teachers. According to the National
Education Association (NEA), teachers generally spend more than 40 hours each week
Teachers
working on instructional duties. The file Teachers contains data on the number of hours
worked per week for a sample of high school science teachers and English teachers.
LO 4, 5
a. Construct a strip chart comparing the number of hours worked per week for high
school science teachers to the number of hours worked per week for English teach-
ers. Be sure to employ techniques to mitigate occlusion.
b. Compute the 25th, 50th, and 75th percentiles of the data for the high school science
teachers and for high school English teachers.
c. Comment on the differences in the number of hours worked per week for high
school science teachers and English teachers.
22. Average Tax Payment Error. A sample of 10,001 Federal income tax returns with
errors has been collected and the amount of payment error for each is provided in the
file FedTaxErrors. A positive value indicates the taxpayer underpaid and a negative
FedTaxErrors
value indicates that the taxpayer overpaid. LO 5, 7
a. What is the sample mean tax payment error made on these erroneous Federal in-
come tax payments?
b. What is the sample standard deviation in payment error amount?
c. What is the margin of error for a 95% confidence interval on the mean payment
error amount?
d. Construct a column chart displaying the sample mean tax payment error with error
bars depicting a 95% confidence interval.
e. Can we conclude with 95% confidence that the mean tax payment error is positive
(a taxpayer underpayment)?
23. Proportion of Tax Underpayment. A sample of 10,001 Federal income tax returns
with errors has been collected and the amount of payment error for each is provided in
the file FedTaxErrors. A positive value indicates the taxpayer underpaid and a negative
FedTaxErrors
value indicates that the taxpayer overpaid. LO 7
a. What is the sample proportion of erroneous income tax returns that correspond to
tax underpayments?
b. What is the margin of error for a 95% confidence interval on the proportion of erro-
neous tax returns corresponding to tax underpayments?
 Problems 225

c. Construct a column chart displaying the sample proportion of tax underpayments

with error bars depicting a 95% confidence interval. Construct the chart to explic-
itly compare the sample proportion and confidence interval to a 50% proportion
benchmark.
d. Can we conclude with 95% confidence that more than 50% of tax payment errors
correspond to tax underpayments?
24. Restaurant Franchising (Revisited). In this problem we revisit the simple linear
regression model from Problem 13 for predicting a restaurant’s quarterly sales with the
size of the nearby student population. The chart and associated data are contained in
Restaurant
FranchiseChart the file RestaurantFranchiseChart. LO 8
a. Modify the chart to convey the uncertainty related to the predictions corresponding
to the simple linear regression model.
b. For approximately what value of student population is the regression model the
most accurate?
25. Forecasting Stock Price (Revisited). In this problem we revisit the chart from Prob-
lem 14 displaying historical price and forecasted price for a stock. The chart and asso-
ciated data are contained in the file DayTradingChart. LO 8
DayTradingChart
a. Modify the chart to convey the uncertainty related to the forecasts of the stock
prices in future days.
b. What can be said about the prediction of the stock price one day in the future versus
the prediction of the stock price 10 days in the future?
Chapter 6
Exploring Data Visually
CONTENTS

DATA VISUALIZATION MAKEOVER 6-4 ANALYSIS OF MISSING DATA

Arctic Sea Ice Volume Types of Missing Data
Exploring Patterns Associated with Missing Data
6-1 INTRODUCTION TO EXPLORATORY DATA 6-5 VISUALIZING TIME SERIES DATA
ANALYSIS Viewing Data at Different Temporal Frequencies
Espléndido Jugo y Batido, Inc. Example Highlighting Patterns in Time Series Data
Organizing Data to Facilitate Exploration Rearranging Data for Visualization
6-2 ANALYZING VARIABLES ONE AT A TIME 6-6 VISUALIZING GEOSPATIAL DATA
Exploring a Categorical Variable Choropleth Maps
Exploring a Quantitative Variable Cartograms
6-3 RELATIONSHIPS BETWEEN VARIABLES SUMMARY
Crosstabulation
GLOSSARY
Association Between Two Quantitative
Variables PROBLEMS

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Organize and arrange data to facilitate data explo- LO 5 Create and compare visualizations that display the
ration and visualization. association between two quantitative variables.
LO 2 Create and interpret visualizations to explore the LO 6 Define correlation and how to estimate its strength
distributions of individual variables. on a scatter chart.
LO 3 Construct crosstabulations and their related LO 7 Define the types of missing data and understand its
charts to explore patterns in data involving implication on how to address it.
multiple variables.
LO 8 Create visualizations for time series data and iden-
LO 4 Select an appropriate chart type for exploring pat- tify temporal patterns.
terns based on type of data and goal of analysis.
LO 9 Explain the strengths and weaknesses of chorop-
leth maps and cartograms for exploring data.
 Data Visualization Makeover 227

D ata V is u ali z ation M a k eover

Arctic Sea Ice Volume

The visualization of data sets can be helpful in gen- meteorology, biology, and ecology of the ice-covered
erating insights from data. However, it can also be regions on Earth. One of the Polar Science Center’s
challenging to choose the best type and format of studies involves the volume of ice covering the Arctic
visualization to best generate insights from data Sea. Figure 6.1 displays a chart similar to that con-
exploration. The Polar Science Center affiliated with structed by Haveland-Robinson Associates that plots
the University of Washington is a custodian of several the monthly average Arctic Sea ice volume from 1979
data sets monitoring the oceanography, climatology, to 2014.

Figure 6.1 Radar Chart of Monthly Ice Volume Observations

As Figure 6.1 shows, this chart corresponds to a lot of of week, months), but years do not repeat, and to add
data points. However, the spherical nature of the chart an additional year’s data to this chart would require
makes it difficult for the audience to quickly determine entirely redrawing the chart (most likely dropping off
insights from the data. Other than facilitating the pithy the oldest year’s worth of data). The chart also heavily
title of “Arctic Death Spiral,” the circular orientation relies on gridlines, which decreases the chart’s data-
of the chart does not lend itself well to displaying ink ratio. Because the months are plotted based on
the data. Often a circular display such as this is used their corresponding ice-volume values, the chart does
when data are repeating over a period of time (days not display months in chronological order as most
(Continued)
228 Chapter 6 Exploring Data Visually

audiences would expect. This makes the choice to of the year’s observations, such as the first and third
show the ice-volume values by month more confusing quartile, the median, and the mean. Perhaps more
than helpful for the audience. importantly, the long-term decrease in ice volume in
An alternative approach to visualizing these data terms of these statistical measures is clearly displayed.
is to remove the individual month values from the An additional benefit of the box and whisker plot
chart. Instead, the analyst could still use the monthly is that data at the most granular level could be used
average data, but present it in a manner that conveys rather than monthly averages. In this case, the Polar
the general pattern in the data, which in this case is Science Center has daily observations of the Arctic
a decrease in the ice volume. Figure 6.2 contains a Sea ice volume. These daily observations can be used
sequence of box and whisker plots each based on as the basis of each year’s box and whisker plot if the
12 monthly averages from a corresponding year. Each analyst determines that the audience is interested in
box and whisker plot provides statistical summaries exploring the variation at the daily level.

Figure 6.2 Annual Box and Whisker Plots of Monthly Ice Volume Observations

In this chapter, we describe the role of data visualization in exploring data. Data are often
said to be “dirty” and “raw” before they have been put in a form that is best suited for
in-depth analysis. We describe how to leverage data visualization during the examination
of missing data and unusual values in the data cleansing process. We demonstrate how to
use different types of charts used to search for patterns within a single variable, such as the
decreasing ice volume discussed in the Data Visualization Makeover. Then we introduce
the use of crosstabulation and scatter charts to investigate patterns between two or more
variables. We conclude the chapter with specific considerations when exploring time series
data and geospatial data.

6-1 Introduction to Exploratory Data Analysis

As a first step in an analytical study, it is critical to examine and explore the data.
This exploratory data analysis (EDA) makes heavy use of descriptive statistics and
visualization to gain an initial understanding of the data. The objectives of EDA include
(1) detection of errors, missing values, and any other unusual observations; (2) characteri-
zation of the distribution of the values for the individual variables; and (3) identification of
patterns and relationships between variables. Visual display is an essential principle of EDA,
as it allows the analyst to translate the information contained in the rows and columns of
data into charts, providing “first looks at the data” that achieve the EDA objectives.
 6-1 Introduction to Exploratory Data Analysis 229

To understand the challenges in exploring data, we consider its structural dimensions.

Tall data occurs when the number of records (rows) is large. Wide data occurs when the
number of variables (columns) is large. As data grow taller or wider, the possibility of data
errors and missing values increases. In addition, wide data becomes increasingly arduous to
explore because there are a large number of possible combinations of variables to examine.

Espléndido Jugo y Batido, Inc. Example

Let us consider an example involving Espléndido Jugo y Batido, Inc. (EJB), a company
that manufactures bottled juices and smoothies. EJB produces its products in five fruit
flavors (apple, grape, orange, pear, and tomato) and four vegetable flavors (beet, carrot,
celery, and cucumber), and it ships these products from distribution centers (DCs) in Idaho,
Mississippi, Nebraska, New Mexico, North Dakota, Rhode Island, and West Virginia.
EJB management has retrieved data on each order it has received over the past three years and
stored it in the file EJB (partially displayed in Figure 6.3). Each record in the data corresponds to
the amount of one product (a combination of category and flavor) included in a single order, so an
order can consist of multiple records in the data. Each record in the data provides the
●● Order ID
●● flavor
●● category (juice or smoothie)

●● dollar amount sold

●● date ordered

●● date delivered

●● DC that filled the order

●● an indication on whether the order was received from a new customer

●● service satisfaction rating reported by the customer

●● product satisfaction rating reported by the customer

FIGURE 6.3 A Portion of Espléndido Jugo y Batido, Inc. Data in the file EJB

EJB

Missing data entries may be A quick scan of the data set shows that there are missing data and that these appear as
coded in a variety of ways, such
blank entries. To visualize the frequency and pattern of missing entries in this data set, we
as blank entries, “NA” entries,
or entries with unrealistic values
execute the following steps.
such as -9999.
Step 1. Select the cell range A1:J19932
Step 2. Click the Home tab on the Ribbon
Click the Conditional Formatting button Formatting
Conditional in the Styles group

Select New Rule…

Step 3. When the New Formatting Rule dialog box appears:
In the Select a Rule Type: area, select Format only cells that contain
In the Edit the Rule Description: area:
In the leftmost box under Format only cells with: select Blanks
Click the Format button, and when the Format Cells dialog box appears:
On the Fill tab, select the black color rectangle from the Back-
ground Color: gallery
Click OK to close the Format Cells dialog box
Click OK to close the New Formatting Rule dialog box
230 Chapter 6 Exploring Data Visually

Figure 6.4 shows the result of the preceding steps. Cells with missing data are colored
black to visually emphasize the frequency and pattern of the missing entries. For the EJB
data, we see that the only variables with missing entries are the Service Satisfaction Rating
and the Product Satisfaction Rating. Presumably, some customers neglect to provide these
voluntary ratings when solicited. Later in this chapter, we will investigate the nature of
these missing data more closely.

FIGURE 6.4 A Portion of the EJB Data Conditionally Formatted to

Display Pattern of Missing Data

Changing the Zoom setting of

the spreadsheet on the View
tab of the Ribbon allows you
to view more or less of the
data set at a time and view
the pattern of missing data at
more or less granular level.

Organizing Data to Facilitate Exploration

In this section, we introduce the functionality of Excel’s Table object to facilitate explo-
ration of a data set. The following steps show how to create a Table object in Excel using
the data in the file EJB.
We execute the steps to Step 1. Click the Insert tab on the Ribbon
create the Excel Table on a
version of the EJB data that Click the Table button in the Tables group
Table
does not have the missing cell
entries highlighted in black. Step 2. When the Create Table dialog box appears:
Enter =$A$1:$J$19932 in the Where is the data for your table? box
Select the My table has headers box
Click OK
Figure 6.5 displays a portion of the table that results from these steps. In addition to the
distinctive banded formatting that an Excel Table has by default, we can recognize that data
in Excel are in a Table object because of the presence of the Table Design tab in the Ribbon
when selecting any cell in the range of the data. By default, the Table containing the data is
named Table1, but we rename the table with the following steps.
 6-1 Introduction to Exploratory Data Analysis 231

Step 3. Select any cell in the range of the data, for example, cell A1
Step 4. Click the Table Design tab on the Ribbon
In the Properties group, enter EJBData in the box below Table Name:

The formatting of an Excel

Table can be easily modified FIGURE 6.5 A Portion of the EJB Data Formatted as Excel Table
on the Table Design tab. The
Table Style Options group
allows the selection of various
options. The Table Styles
group provides a gallery of
table format templates from
which to select.

As the Filter Arrow next to each column heading in Figure 6.5 suggests, when an
Excel Table is created, the data set is automatically prepared to filter with respect to a vari-
able’s value. To demonstrate this functionality, suppose we are interested only in viewing
the records corresponding to beet smoothies. To filter the data to list only this flavor and
category combination, we execute the following steps.
Step 1. Click the Filter Arrow in cell B1 next to Flavor
Deselect the box next to (Select All)
Select the box next to Beet
Click OK
Step 2. Click the Filter Arrow in cell C1 next to Category
Deselect the box next to (Select All)
Select the box next to Smoothies
Click OK
The result is a display of only the records that were part of orders for beet smoothies (see
Figure 6.6).

FIGURE 6.6 A Portion of the EJB Data Filtered to Show Only Beet
Smoothie Records

Suppose now that we only want to view beet smoothie records that were greater than or equal to
$1,500 in sales. To filter on a quantitative variable such as $ Sales, we execute the following steps.
Step 1. Click the Filter Arrow in cell D1, next to $ Sales
Step 2. Select Number Filters and then Greater Than Or Equal To…
Step 3. When the Custom AutoFilter dialog box appears:
In the top box on the left in the Show rows where: area, select is greater
than or equal to and enter 1500 in the adjacent box
Click OK
232 Chapter 6 Exploring Data Visually

The result is a display of only the beet smoothie records of at least $1,500 in sales
(see Figure 6.7).

FIGURE 6.7 A Portion of the EJB Data Filtered to Show Only Beet
Smoothie Records with Sales $ $1500

We can further filter the data by choosing the Filter Arrow in other columns. We can
make all the data visible again by clicking the Filter Arrow in each filtered column (as
denoted by the icon) and checking (Select All). We can also undo all filtering by click-
ing Clear in the Sort & Filter group of the Data tab.
In addition to filtering records, the Filter Arrow for a column can be used to quickly sort
data according to magnitude (for a quantitative variable), alphabetically (for a categorical
variable), or chronologically (for a variable coded as a date). After removing all previous
filters and making all data visible again, the following steps demonstrate how to sort a Table
according to the Date Delivered variable chronologically from the oldest date to the newest
date (Figure 6.8 shows a portion of the resulting table).
Step 1. Click the Filter Arrow in column F1 next to Date Delivered
Step 2. Select Sort Oldest to Newest

FIGURE 6.8 A Portion of the EJB Data Sorted Oldest Date Delivered to
Newest Date Delivered

An Excel Table enables the creation of a new variable that is calculated from other vari-
ables in the data set. To demonstrate, the following steps add a new variable to the Table
containing the EJB data called Time to Deliver.
Step 1. Click the top of column G, right-click and select Insert from the
dropdown menu
Step 2. In cell G1, enter Time to Deliver
Step 3. In cell G2, enter =F2-E2
Step 4. Select column G by clicking the top of the column
Step 5. In the Home tab of the Ribbon, within the Number group:
From the drop-down menu, select Number
Click the Decrease Decimal 000 button twice
Step 6. In the Home tab of the Ribbon, within the Alignment group:
Click the Align Right button
Figure 6.9 shows the result of these steps. When pressing Enter to execute the formula in
Step 3 above, entries for the entire column are automatically completed with the analogous
calculations.
 6-1 Introduction to Exploratory Data Analysis 233

FIGURE 6.9 A Portion of the EJB Data after Insertion of Time to Deliver
Variable

If you append a record to a row adjacent to an Excel Table or add a variable in a column
adjacent to an Excel Table, the Table is automatically resized to include this new row or col-
umn. To ease analysis of the EJB data based on order date, it may help to decompose the Date
Ordered variable into three new variables: Year Ordered, Month Ordered, and Day Ordered.
The following steps create these new variables in the Excel Table (shown in Figure 6.10).
Step 1. In cell L1, enter Year Ordered
Step 2. In cell M1, enter Month Ordered
Step 3. In cell N1, enter Day Ordered
Step 4. In cell L2, enter =YEAR(E2)
In the Excel function TEXT, the Step 5. In cell M2, enter =TEXT(E2,“mmm”)
second argument specifies Step 6. In cell N2, enter =DAY(E2)
the format of value (date)
to be converted to text.
TEXT(E2, "mmm") returns
FIGURE 6.10 A Portion of the EJB Data after Insertion of Year Ordered,
the abbreviated month
corresponding to the date Month Ordered, and Day Ordered Variables (Columns H
listed in cell E2. through K Hidden

An Excel Table supports the automatic calculation of summary statistics for each col-
umn (variable). For example, suppose we are interested in computing the average value of
$ Sales for a set of filtered observations and comparing that to the average value of $ Sales
for all observations. The following steps demonstrate this comparison for beet smoothies
starting with the entire (unfiltered) table of observations in the file EJB.
Step 1. Select any cell in the range of the Table, for example, cell A2.
Step 2. In the Table Design tab on the Ribbon, select Total Row in the Table Style
Options group.
Step 2 will append a row to the bottom of the Table with the label Total in the first column.
The following steps execute the calculation of the average value of $ Sales.
Step 3. In the Total row (row 19933), select the entry corresponding to the $ Sales column
Inserting the Total Row in a
Table automatically computes
Click the Drop-down Menu Arrow and select Average from the menu
the Sum of the variable in the
Figure 6.11 shows that $894.18 is the average sales amount when considering all records.
last column. This calculation
can be removed by clicking on
A convenient feature of an Excel Table is that this Total calculation updates appropriately
the dropdown menu arrow in for filtered records. Figure 6.12 shows that $706.05 is average sales amount for records
this cell and selecting None. corresponding to beet smoothies.
234 Chapter 6 Exploring Data Visually

FIGURE 6.11 Calculation of Average $ Sales for All Records Using Table’s
Total Row

An Excel Table accommodates

tall data by automatically FIGURE 6.12 Calculation of Average $ Sales for Beet Smoothie Records
replacing the column labels Using Table’s Total Row
of A, B, C,… with the column
headers in the first row
(variable names) as you
scroll down.

In the next two sections, we will demonstrate how an Excel Table facilitates EDA anal-
ysis and the construction of charts. When relevant and possible, we will perform the statis-
tical summaries and visualization of EDA with a PivotTable and PivotChart. A PivotTable
is an Excel tool that summarizes data for one or more variables. A PivotChart is an Excel
charting tool connected with a PivotTable.

N otes 1 C omments

An Excel Table automatically names cell ranges using the col- the named range when selecting the cell range for the cell
umn headers so that cells can be referenced with these names formula. Named ranges make Excel more dynamic by allowing
rather than cell column-row references (for example, C7). If a formula to reference ranges that are not fixed to column-
creating a cell formula by clicking on the cells involved, these row references that may change if new data are added to the
cells in the Excel Table will automatically be referenced using spreadsheet.

6-2 Analyzing Variables One at a Time

Chapter 5 discusses the As a general principle, it is almost always a good idea to first examine one variable at
various charts used to analyze a time before exploring relationships between two or more variables. Examining one
the distribution of a variable’s
variable (column) of a data set is called univariate analysis. Univariate analysis focuses
values and provides step-by-
step instructions on how to
on the distribution of a variable’s values. In this section, we discuss how to examine
construct them. categorical and quantitative variables.

Exploring a Categorical Variable

We demonstrate the construction of charts displaying an individual categorical variable’s
distribution using Excel’s PivotChart functionality. We will use the data from the EJB
example formated using an Excel Table found in the file EJBTable.
 6-2 Analyzing Variables One at a Time 235

Step 1. Select any cell in the range of the data, for example, cell A3
Step 2. Click the Insert tab on the Ribbon
In the Charts group, click the PivotChart button
EJBTable PivotChart
Step 3. When the Create PivotChart dialog box appears:
Under Choose the data that you want to analyze, choose Select a table
or range and enter EJBData in the Table/Range: box
Under Choose where you want the PivotChart to be placed, select
New Worksheet
Click OK
The resulting initial (empty) PivotTable and PivotChart are shown in Figure 6.13. With the
PivotChart selected, the PivotChart task pane is activated. Each of the 14 columns (vari-
ables) is identified as a PivotChart Field by Excel. PivotChart Fields may be chosen to rep-
resent axes (categories), legends (series), filters, or values in a PivotChart. The following
steps show how to use Excel’s PivotChart Field List to assign Flavor to the horizontal axis
and chart the percent frequency of Order IDs for each flavor.
When the PivotTable is Step 4. In the PivotChart Fields task pane, under Choose fields to add to report:
selected, the PivotTable task
Drag the Flavor field to the Axis (Categories) area under Drag fields
pane will be activated instead
of the PivotChart task pane.
between areas below:
The PivotTable task pane is Drag the Order ID field to the Values area
similar to the PivotChart task Step 5. Click Sum of Order ID in the Values area under Drag fields between areas
pane with the only differences below:
being that a PivotTable has
Step 6. Select Value Field Settings… from the list of options
Rows and Columns areas
instead of Axis (Categories)
Step 7. When the Value Field Settings dialog box appears:
and Legend (Series) areas. Click the Summarize Values By tab and under Summarize value field
by, select Count
Click the Show Values As tab and in the Show values as dropdown
menu, select % of Grand Total
Click OK
Step 8. Click any of the columns in the PivotChart. While the columns are selected,
right-click a column, then select Sort and Sort Largest to Smallest

FIGURE 6.13 Initial PivotTable and PivotChart for the EJB Data
236 Chapter 6 Exploring Data Visually

Further editing will result in a chart that matches Figure 6.14. Figure 6.14 shows the
percent frequency of records by flavor. We observe that orange is the most commonly
ordered flavor (15.86% of all records) and tomato is the least commonly ordered flavor
(6.20% of all records).
Similar frequency charts can be constructed for the other categorical variables in the
EJB data (Category, Distribution Center, New Customer) for which there is no natural
ordering of the categories.

FIGURE 6.14 PivotTable and PivotChart for Flavor Distribution of Records

in the EJB Data

In the EJB data, Service Satisfaction Rating and Product Satisfaction Rating are ordinal
variables—categorical variables with a natural ordering. Frequency charts for ordinal vari-
ables can be constructed in a similar manner as frequency charts for general categorical
variables, but one must be careful not to disrupt the natural ordering of the variable values
with sorting. For example, Figure 6.15 displays the distribution of records by the Service
Satisfaction Rating. Because there is a natural ordering of the values (1 is the lowest rat-
ing, 2 is the second lowest, etc.), it would be inappropriate to sort these values in order
of increasing or decreasing frequency. Figure 6.15 shows that the most common service
satisfaction rating was a 5, but that also 27.47% of the records were missing a service
satisfaction rating response. In an upcoming section, we discuss ways to address missing
data such as these.
 6-2 Analyzing Variables One at a Time 237

FIGURE 6.15 PivotTable and PivotChart for Service Satisfaction Rating

Distribution of Records in the EJB Data

Exploring a Quantitative Variable

We demonstrate the construction of charts displaying an individual quantitative variable’s dis-
tribution using Excel’s PivotTable and PivotChart functionality. We assume that a PivotTable
and PivotChart using the data in the Table named EJBData have previously been created
(corresponding to the analysis in Figure 6.14). In the following steps, we outline how to con-
struct another PivotTable and PivotChart on the same source data (the Excel Table EJBData).
Step 1. Select the worksheet containing the previously constructed PivotTable and Pivot-
Chart and make a copy of this worksheet by right-clicking on the worksheet name,
selecting Move or Copy, and when the Move or Copy dialog box appears:
Select the check box for Create a copy
Click OK
If the PivotChart Fields task Step 2. Click the PivotChart in the new worksheet
pane does not appear when When the PivotChart Fields task pane appears, clear the contents of the
you select the PivotChart,
right click the chart and
current PivotTable and PivotChart by deselecting the boxes next to the
choose Show Field List. currently selected fields
The following steps show how to use Excel’s PivotChart Field List to assign $ Sales to
the horizontal axis and chart the percent frequency of Order IDs in the bins specified by a
range of $ Sales values.
Step 3. In the PivotChart Fields task pane, under Choose fields to add to report:
Drag the $ Sales field to the Axis (Categories) area
Drag the Order ID field to the Values area
Step 4. Click Sum of Order ID in the Values area
238 Chapter 6 Exploring Data Visually

Step 5. Select Value Field Settings... from the list of options

Step 6. When the Value Field Settings dialog box appears:
Click the Summarize Values By tab and under Summarize value field
by, select Count
Click the Show Values As tab and in the Show values as drop-down
menu select % of Grand Total
Click OK
Step 7. Right-click in cell A2 or any other cell containing a $ Value row label
Select Group from the list of options
Step 8. When the Grouping dialog box appears:
Enter 0 in the Starting at: box
Enter 2000 in the Ending at: box
Enter 100 in the By: box
Click OK
Further editing will result in a chart that matches Figure 6.16. Figure 6.16 displays the
distribution of sales amounts at the product record level. We see that the records with
smaller sales amounts are slightly more common and that the frequency of records
with increasingly larger sales amounts gradually decreases.

FIGURE 6.16 PivotTable and PivotChart for Distribution of $ Sales at the

Product Record Level in the EJB Data

In a similar manner, we can visualize the distribution of the Time to Deliver variable
at the product record level. Figure 6.17 displays the corresponding PivotTable and
PivotChart. We observe that the delivery time ranged from 2 days to 21 days, with a
most common delivery time of 3 days. The distribution of delivery times is skewed to
the right by rare lengthy times to delivery.
 6-2 Analyzing Variables One at a Time 239

FIGURE 6.17 PivotTable and PivotChart for Distribution of Time to Deliver at the Product
Record Level in the EJB Data

Another way to visualize the distribution of a quantitative variable is the box and
whisker plot. The box and whisker plot summarizes the values of a variable by displaying
various statistical measures: first quartile, second quartile (median), third quartile, mean,
as well as interquartile range (IQR), which is computed as the difference between the third
quartile and the first quartile. Because the box and whisker plots use the median to measure
the central location of a variable and the IQR to measure the deviation of a variable, their
display is robust to the effects of extreme values, making them a valuable EDA visualiza-
tion. The box and whisker plot is good at presenting information about a variable’s central
tendency, the symmetry or lack of symmetry in the distribution, as well as data points with
extreme values.
For example, Figure 6.18 displays a box and whisker plot for the $ Sales variable.
The first quartile of the $ Sales variable is depicted by the horizontal line forming the
bottom of the box at a height of $401.02 on the vertical axis. The second quartile (the
median) of the $ Sales variable is depicted by the horizontal line on the inside of the
box at a height of $851.74 on the vertical axis. The third quartile of the $ Sales variable
is depicted by the horizontal line forming the top of the box at a height of $1,355.08
on the vertical axis. The mean of the $ Sales variable is depicted by the X on the inside of
the box at a height of $894.18. The vertical lines extending from the top and bottom
of the box are called whiskers. The top whisker extends to the largest value of $ Sales
that is less than or equal to

third quartile + 1.5 × (third quartile − first quartile)

= 1355.08 + 1.5 × (1355.08 − 401.02) = $2786
240 Chapter 6 Exploring Data Visually

Because the largest value of $ Sales is $1,999.92, which is less than $2,786, the top whis-
ker extends only to $1,999.92. The bottom whisker extends to the smallest value of $ Sales
that is greater than or equal to

first quartile − 1.5 × (third quartile − first quartile) = 401.02 − 1.5 × (1355.08 − 401.02)
= −$1030.07

Because the smallest value of $ Sales is $0.05, which is greater than −$1,030.07, the bot-
tom whisker extends only to $0.05.
Box and whisker plots From the box and whisker plot of $ Sales in Figure 6.18, we can deduce that the distri-
are discussed in detail in bution of $ Sales has a slight positive skew as the upper whisker is just a bit longer than the
Chapter 5. Box and whisker
lower whisker. The mean being slightly larger than the median reinforces the implication
plots cannot be constructed
through the PivotChart tool. of slight positive skew. All of these observations can be confirmed by referring to the histo-
gram of the $ Sales variable in Figure 6.16.

FIGURE 6.18 Box and Whisker Plot of $ Sales at the Product Record
Level in the EJB Data

Now consider the box and whisker plot for the Time to Deliver variable in Figure 6.19.
We observe that the mean delivery time is larger than the median delivery time.
Additionally, the bottom whisker is much shorter than the top whisker, suggesting that
below there is a high concentration of delivery time observations over a narrow range of
small values. The longer top whisker indicates delivery time observations above the median
are spread out over a relatively wide range. The short lower whisker, long upper whisker,
and mean value larger than the median value suggest that the delivery time distribution is
With respect to a box and
whisker plot, an outlier is
positively skewed. Additionally, the presence of several outliers beyond the top whisker
an observation that is more suggest that the delivery time distribution has a long tail of relatively large values. It is a
extreme than the lower or good idea to inspect the records corresponding to these outliers to confirm that these are
upper whisker. However, there accurately reported and not the result of an error. If the value of an outlier is the result of
is no universal definition for
an error or if the observation occurred in a circumstance that makes it inappropriate for an
what constitutes an outlier in
a data set. Therefore, different
analytical study, the observation may be removed from consideration. However, removing
software packages may define outlier observations without warrant can distort analysis by artificially reducing the varia-
outliers slightly differently. tion in a variable.
 6-2 Analyzing Variables One at a Time 241

FIGURE 6.19 Box and Whisker Plot of Time to Deliver at the Product
Record Level in the EJB Data

N otes 1 C omments

1. Excel comes bundled with a Data Analysis add-in that auto- To visualize the distribution of $ Sales at the order level,
mates several statistical procedures. After activating this we first must use a PivotTable to aggregate the data appro-
add-in, Data Analysis can be found in the Analysis group priately. Specifically, a PivotTable with Order ID in the Rows
in the Data tab in the Ribbon. By selecting Data Analysis area and $ Sales in the Values area with Sum selected as
and then Descriptive Statistics, the statistical summaries of the Value Field Setting will result in a table of data listing
the quantitative variables can be calculated. each order with its corresponding sales amount. This data
2. If Data Analysis does not appear in the Analysis group in can then be copied and pasted outside the PivotTable and
the Data tab, you will have to load the Analysis Toolpak used as the basis for a histogram and box and whisker chart
add-in into Excel. To do so, click the File tab in the Ribbon (see Figures (a) and (c) below).
and click Options. When the Excel Options dialog box To visualize the distribution of Time to Deliver at the
appears, click Add-ins from the menu. Next to Manage:, order level, we first must use a PivotTable to aggregate the
select Excel Add-ins, and click Go... at the bottom of the data appropriately. Specifically, a PivotTable with Order ID
dialog box. When the Add-Ins dialog box appears, check in the Rows area and Time to Deliver in the Values area
the box next to Analysis Toolpak and click OK. with Average selected as the Value Field Setting will result
3. The distribution analysis of the $ Sales and Time to Deliver in a table of data listing each order with its corresponding
variables in Figures 6.16, 6.17, 6.18, and 6.19 is based sales amount. Because all products in an order are deliv-
on record-level data. However, because some product ered at the same time, the average time to deliver reflects
records may be part of the same order (and share the the observed delivery time. This data can then be copied
same Order ID), it also may be insightful to visualize the and pasted outside the PivotTable and used as the basis
distribution of the $ Sales and Time to Deliver variables for a histogram and box and whisker chart (see Figures (b)
at the order level. and (d) below).
(Continued)
242 Chapter 6 Exploring Data Visually

6-3 Relationships between Variables

After satisfactorily examining one variable at a time, the natural progression is to examine
relationships between variables. Examining two or more variables at a time to explore rela-
tionships is called multivariate analysis. Generally, multivariate analysis involves pairs of
variables (and thus is bivariate) but may involve three or more variables. Whether the vari-
ables of interest are categorical or quantitative dictates the statistical summary and visual-
ization technique deployed. In this section, we first discuss a crosstabulation, which is a
tabular summary of the statistical measure of a variable of interest with respect to two or
more (typically categorical) variables. Then we discuss the use of scatter charts to visualize
the relationship between two quantitative variables.

Crosstabulation
In this section, we will demonstrate crosstabulation analysis of two or more variables and
associated visualizations using Excel’s PivotTable and PivotChart functionality. Suppose
EJB is interested in investigating how the average sales amount of a record depends on the
DC from which it was shipped and its product category. The following steps show how to
construct a crosstabulation of Distribution Center as the variable in the table rows and Cat-
egory (juices or smoothies) as the variable in the table columns. $ Sales is the variable to be
summarized with respect to its average value in the corresponding cross-sections of Distri-
bution Center and Category. We begin our steps from an empty PivotTable and PivotChart
using the Excel Table EJBData as the source data as in Figure 6.13.

To clear the content of

Step 1. Select any cell in the range of the empty PivotTable, for example, cell A5
the PivotChart, click the Step 2. When the PivotTable Fields task pane appears, under Choose fields to add to
PivotChart. From the report:
PivotChart Fields task pane, Drag the $ Sales field to the Values area
clear the contents of the Drag the Distribution Center field to the Rows area
current PivotTable and
PivotChart by deselecting the
Drag the Category field to the Columns area
boxes next to the currently Step 3. Click Sum of $ Sales in the Values area
selected fields. Step 4. Select Value Field Settings… from the list of options
6-3 Relationships between Variables 243

Step 5. When the Value Field Settings dialog box appears:

Click the Summarize Values By tab and under Summarize value field
by, select Average
Click OK
Further editing will result in a chart that matches Figure 6.20. Each entry in the PivotTable of
Figure 6.20 corresponds to the average dollar sales per record computed over a corresponding
cross-section of the data. For example, the value of 874.01 in cell B6 means that juice records
shipped from the DC in Nebraska (NE) had an average sales amount of $874.01. The Grand
Total amounts at the end of each column or row correspond to the average sales amounts over
the corresponding column or row. For example, the value of 865.78 in cell D5 means that
records shipped from the DC in North Dakota (ND) had an average sales amount of $865.78
(computed over both product categories). Analogously, the value of 887.19 in cell C10 means
that smoothie records had an average sales amount of $887.19 (computed over all DCs).
From Figure 6.20, we observe several patterns in the average sales amount of a record
relative to the DC and product category. First, the average sales amount for a juice
record is larger than the average sales amount for a smoothie record for all DCs except
Nebraska and West Virginia. The largest average sales amounts for both juices and
smoothies are associated with records shipped from Rhode Island. The smallest average
sales amount is associated with smoothie records shipped from Nebraska.

FIGURE 6.20 PivotTable and PivotChart of Crosstabulation of Average $

Sales over Distribution Center and Product Category

While the variables used in the rows and columns of a PivotTable are typically cate-
gorical variables, it is possible to place a quantitative variable in the rows or columns of a
PivotTable and then form bins by grouping consecutive values of the quantitative variable.
Suppose EJB is interested in investigating comparing how the distribution of the sales
amount of a product record depends on whether a customer is new or existing.
244 Chapter 6 Exploring Data Visually

The following steps show how to construct a crosstabulation of $ Sales as the variable in the
table rows and New Customer as the variable in the table columns. As we are interested in
the frequency distribution of records over the cross-section of $ Sales and New Customer,
we may select and count over any variable that is not missing values. We choose Order ID
as the variable to be summarized with respect to its count value in the corresponding cross-
sections of Distribution Center and Category. We begin our steps from an empty PivotTable
and PivotChart using the Excel Table EJBData as the source data (as in Figure 6.13).
Step 1. Select any cell in the range of the empty PivotTable, for example, cell A5
Step 2. When the PivotTable Fields task pane appears, under Choose fields to add
to report:
Drag the Order ID field to the Values area
Drag the $ Sales field to the Rows area
Drag the Category field to the Columns area
Step 3. Click Sum of Order ID in the Values area
Select Value Field Settings… from the list of options
Step 4. When the Value Field Settings dialog box appears:
Click the Summarize Values By tab and under Summarize value field
by, select Count
Click the Show Values As tab and in the Show values as drop-down
menu select % of Column Total
Click OK
Step 5. Right-click in cell A5 or any other cell containing a $ Value row label
Select Group from the list of options
Step 6. When the Grouping dialog box appears:
Enter 0 in the Starting at: box
Enter 2000 in the Ending at: box
Enter 100 in the By: box
Click OK
Step 7. With the PivotChart selected, right-click and select Change Chart Type…
In Change Chart Type dialog box, select Line
Click OK
Further editing will result in a chart that matches Figure 6.21. Each entry in the PivotTable
of Figure 6.21 corresponds to the percentage of the total number of records in a column that
occur in a corresponding cross-section of the data. For example, the value of 6.54% in cell
B4 means that 6.54% of all records corresponding to existing customers have sales amounts
in the range (100, 200] (greater than $100 and less than or equal to $200). From Figure 6.21,
we observe that for both existing and new customers, the frequency of records gradually
decreases as the sales amount of the record increases. It does not appear that the customer
status (existing or new) dramatically affects the sales amount of an record.
A PivotTable can be used to consider a crosstabulation of more than two variables. Suppose
EJB is interested in examining total annual sales patterns over the years of 2018, 2019, and
2020, while also considering whether records originated from new customers and the DC from
which the records were shipped. The following steps show how to construct a crosstabulation
of Distribution Center and Year Ordered as the variables in the table rows and New Customer
as the variable in the table columns. $ Sales is the variable to be summarized with respect to its
sum value in the corresponding cross-sections. We begin our steps from an empty PivotTable
and PivotChart using the Excel Table EJBData as the source data (as in Figure 6.13).
Step 1. Select any cell in the range of the empty PivotTable, for example, cell A5
Step 2. When the PivotTable Fields task pane appears, under Choose fields to add
to report:
Drag the $ Sales field to the Values area
Drag the Distribution Center field to the Rows area
Drag the Year Ordered field to the Rows area
Drag the New Customer? field to the Columns area
 6-3 Relationships between Variables 245

Step 3. Click the PivotChart, right-click and select Change Chart Type... from the menu
In the Change Chart Type dialog box, select Column from the list of
charts and then select Stacked Column from the gallery
Click OK
You should confirm that The next steps add an Excel feature called Slicers. An Excel Slicer provides a visual
$ Sales is being summarized
method for filtering the data considered by the PivotTable and PivotChart.
by the Sum function as
indicated by Sum of $ Sales Step 4. With the PivotChart selected, in the Insert tab in the Ribbon, select Slicer in
in the Values area of the
the Filters group
PivotTables Fields task pane.
In the Insert Slicers dialog box, select Distribution Center, New Cus-
tomer?, and Year Ordered
Click OK

FIGURE 6.21 PivotTable and PivotChart of Crosstabulation of Counting

Order ID over $ Sales and New Customer
246 Chapter 6 Exploring Data Visually

The next two steps format the Slicers and PivotChart.

Step 5. With the Slicer box for Distribution Center selected, in the Slicer tab in the
Ribbon:
In the Buttons group, enter 7 in the Columns: box
Step 6. With the Slicer box for Years Ordered selected, in the Slicer tab in the Ribbon:
In the Buttons group, enter 3 in the Columns: box
Step 7. With the Slicer box for New Customer? selected, in the Slicer tab in the Ribbon:
In the Buttons group, enter 2 in the Columns: box
Step 8. With the PivotChart selected, in the PivotChart Analyze tab in the Ribbon:
In the Show/Hide group, click the down arrow next to Field Buttons Field
Buttons

Select Hide All

Step 9. Select the vertical axis labels, right-click and select Format Axis… , and in
the Format Axis task pane:
Select Axis Options and in the drop-down menu next to Display units:
Select Thousands
Deselect the check box next to Show display units label on chart
After positioning the Slicer boxes adjacent to the PivotChart and sizing all boxes to create
a seamless display, a single Excel object can be created with the following step. Further
editing will result in a chart that matches Figure 6.22.
Step 10. Holding the Ctrl key, select the three Slicer boxes and PivotChart
Right-click any of the selected objects and select Group
We can remove the borders from a Slicer box by executing the following steps to create a
new slicer style which we can apply to any Slicer in the Excel workbook.
Step 11. With a Slicer box selected, click the Slicer tab on the Ribbon
Step 12. In the Slicer Styles group, right-click the current style and select Duplicate

FIGURE 6.22 PivotChart for Total Sales with Slicers for Distribution Center,
Year Ordered, and New Customer

Total Sales (Thousand $) by Distribution Center, Year Ordered,

and New Customer
1400
1200
1000 New
800 Customer?
Yes
600
No
400
200
0
2018
2019
2020
2018
2019
2020
2018
2019
2020
2018
2019
2020
2018
2019
2020
2018
2019
2020
2018
2019
2020

ID MS ND NE NM RI WV
Distribution Center

ID MS ND NE NM RI WV

Year Ordered New Customer?

2018 2019 2020 No Yes

 6-3 Relationships between Variables 247

Step 13. When the Modify Slicer Style dialog box appears, select Whole Slicer from
the Slicer Element: box and click Format
In the Format Slicer Element dialog box:
Click the Border tab and select None in the Presets area
Click OK
Click OK to close the Modify Slicer Style dialog box
Selecting Slicer boxes and applying the newly defined slicer style to these will remove the
borders from them. Further editing will result in a chart that matches Figure 6.22.
Recall that you can select The PivotChart in Figure 6.22 will allow the user to graphically display total dollar sales
multiple items in a slicer by for any combination of years ordered, DCs, and new/existing customer status. For example,
clicking a slicer button while
suppose we want to compare sales of new customers for the New Mexico and Rhode Island
holding the Ctrl key, then
clicking on each additional DCs in 2019 and 2020. We can create a chart for this purpose by selecting NM and RI in
item you wish to select in the Distribution Center slicer, 2019 and 2020 in the Year Ordered slicer, and Yes in the
that slicer. New Customer? slicer. This produces the chart in Figure 6.23.

FIGURE 6.23 PivotChart for Total Sales to New Customers Shipped from
New Mexico and Rhode Island in 2019 and 2020

Association between Two Quantitative Variables

Thus far, we have demonstrated how to use a PivotTable and PivotChart to track a sum-
mary statistic of variable of interest over a crosstabulation of two or more categorical
variables. Often a manager or decision maker is interested in the relationship between two
quantitative variables. In this section, we consider methods for examining the association
between two quantitative variables.
To demonstrate, we consider data for each of New York City’s 55 sub-boroughs (a desig-
nation of a community within New York City). Figure 6.24 contains a partial listing of these
data, which tracks median monthly rent, percentage of college graduates, poverty rate, mean
commute time to workplace, and borough. Suppose we want to examine the relationship
between these different variables. A scatter chart is a useful graph for analyzing the rela-
tionship between two quantitative variables. Figure 6.25 displays scatter chart of Percentage
248 Chapter 6 Exploring Data Visually

College Graduates versus Median Monthly Rent. To clearly convey meaningful patterns, the
aspect ratio and quantitative scales must be appropriately set. With regard to the aspect
ratio, it is generally a good idea to set the height and the width of a scatter chart to be the
same to create a square display that does not favor either variable. When setting the quan-
titative scales for a scatter chart, it is generally recommended to begin each axis at a value
slightly smaller than the smallest value of the corresponding variable and end each axis at a
value slightly larger than the largest value. To add the linear trendline to the scatter chart in
Figure 6.25 and then format the scatter chart, we executed the following steps.
Step 1. Select the data series by clicking on the data points.
Step 2. Right-click and select Add Trendline…
In the Format Trendline task pane, under Trendline Options , select
Linear
Step 3. With the chart selected, click the Format tab in the Ribbon
In the Size group, enter 5 in the Height: box and enter 5 in the Width: box
Step 4. Click the horizontal axis, right-click and select Format Axis…
In the Format Axis task pane, under Axis Options, enter 0 in the
Minimum Bound and enter 2000 for the Maximum Bound
Step 5. Click the vertical axis, right-click and select Format Axis…
In the Format Axis task pane, under Axis Options, enter 0 in the
Minimum Bound and enter 80 for the Maximum Bound

FIGURE 6.24 A Portion of the New York City Sub-Borough Data

NYC

We observe from Figure 6.25 that generally, as median monthly rent increases in a sub
-borough, the percentage of the sub-borough’s inhabitants that are college graduates also
increases. The positive slope of the linear trendline indicates there is a positive correlation
between Percentage College Graduates and Median Monthly Rent. Correlation is a statistical
measure of the strength of the linear relationship between variables. Values of correlation range
between 21 and +1. Correlation values near 0 indicate no linear relationship exists between the
two variables. The closer a correlation value is to +1, the closer the data points on the scatter
chart of the two variables resembles a straight line that trends upward to the right (positive slope).
The closer a correlation value is to 21, the closer the data points on the scatter chart of the two
variables resembles a straight line that trends downward to the right (negative slope). We can cal-
culate the correlation in the data for two variables using the Excel function CORREL. For exam-
ple, the correlation of 0.87 between Percentage College Graduates and Median Monthly Rent is
computed by placing the formula =CORREL(C2:C56, B2:B56) in a cell.
The positive correlation between Percentage College Graduates and Median Monthly
Rent does not imply that a change in one variable causes a change in the other variable,
only that generally speaking an increase (decrease) in one variable generally corresponds
to an increase (decrease) in the other variable. The strength of this positive correlation can
be visually gauged by how close the data points are clustered around the linear trendline.
As a collection, the less total vertical distance between the data points and the trendline,
the stronger the correlation between the two variables. While the sign (positive or negative)
of the correlation is depicted by the slope (positive or negative) of the linear trendline, the
6-3 Relationships between Variables 249

strength of the correlation between two variables is not related to the steepness of the slope
of the linear trendline. The slope reflects the unit change in the variable on the vertical axis
given a unit change in the variable on the horizontal axis. Thus, while the slope is affected
by the units in which the variables are expressed, the correlation is not.
In general, when two variables are correlated, this can either indicate a causal relation-
ship in which one variable causes the other variable’s behavior or it can be the result of a
spurious relationship in which there is no cause-and-effect between the two variables. A
spurious relationship between two variables can arise when (1) both variables are affected
by a third variable, called a lurking variable, (2) the data are biased and not a representa-
tive sample, or (3) the data are insufficient to distinguish it from random coincidence.

FIGURE 6.25 Scatter Chart of Percentage College Graduates versus

Median Monthly Rent with Linear Trendline

Positive Relationship between Rent and

Percentage College Graduates in a Sub-borough
College Graduates (%)
80

70

60 Correlation = +0.87

50

40

30

20

10

0
0 500 1000 1500 2000
Median Monthly Rent ($)

In Figure 6.26, we observe a negative relationship between the poverty rate of a sub-
borough and the percentage of the sub-borough’s inhabitants that are college graduates.
Generally, as poverty rate increases in a sub-borough, the percentage of the sub-borough’s
inhabitants that are college graduates decreases. As depicted by the negative slope of the
linear trendline fit to these data, there is a negative correlation between Percentage College
Graduates and Poverty Rate.
In Figure 6.27, we observe that the data points do not fit closely to the linear trendline
and the correlation is near zero. This means that there is no linear relationship between
Commute Time and Poverty Rate. For these data, there does not appear to be any type of
relationship between Commute Time and Poverty Rate. However, we caution that a near-
zero correlation only means that there is no evidence for a linear relationship between two
variables, but that there may be a nonlinear relationship between them.
To underscore the importance of examining the scatter charts of pairs of quantitative vari-
ables rather than just computing tables of correlation measures, we consider the scatter chart
in Figure 6.28. Figure 6.28 displays data for monthly heating/cooling bills and the mean high
temperature. The data points do not fit closely to the linear trendline and the correlation is near
zero, suggesting there is no linear relationship between the monthly heating/cooling bill and
250 Chapter 6 Exploring Data Visually

FIGURE 6.26 Scatter Chart of Percentage College Graduates versus Poverty Rate with Linear Trendline

Negative Relationship between Poverty and

Percentage College Graduates in a Sub-borough
College Graduates (%)
80
70
60 Correlation = –0.56
50
40
30
20
10
0
0 5 10 15 20 25 30 35 40 45
Poverty Rate (%)

FIGURE 6.27 Scatter Chart of Mean Commute Time versus Poverty Rate with Linear Trendline

No Relationship between Poverty and Commute

Time in a Sub-borough
Commute Time (minutes)
50
45
40
35
30 Correlation = +0.01
25
20
15
10
5
0
0 10 20 30 40
Poverty Rate (%)

the mean high temperature. However, it would be incorrect to conclude there is no relationship
between these two variables. As the nonlinear trendline outlines, there is strong visual evidence
of a nonlinear relationship between these two variables. That is, reading the chart from left to
right, we can see that as the mean high temperature increases, the monthly bill first decreases as
less heating is required and then increases as more cooling is required.
 6-3 Relationships between Variables 251

FIGURE 6.28 Example of Nonlinear Relationship Producing a Near Zero

Correlation

Nonlinear Relationship between Temperature and

Heating/Cooling Bill
Monthly Heating/Cooling Bill ($)
350

The nonlinear trendline in 300

Figure 6.28 was added by
selecting Add Trendline 250
and then specifying the
Polynomial trendline option 200
with Order specified as 2.
150
Correlation = +0.01
100

50

0
0 25 50 75 100 125
Average High Temperature (degrees Fahrenheit)

Consideration of a categorical variable can be added to a scatter chart displaying the

relationship between two quantitative variables. Let us reconsider the relationship between
a sub-borough’s percentage of college graduates and its median monthly rent that was
displayed in Figure 6.25. In Figure 6.25, all 55 observation pairs of Median Monthly Rent
and Percentage College Graduates were plotted as a single data series. To add information
about the borough to the observation pairs of Median Monthly Rent and Median Monthly
Rent, we plot the observation pairs associated with different values of the Borough variable
as different data series. We demonstrate this process with the following steps.
Step 1. Select cells A1:F56
Step 2. Click the Data tab on the Ribbon
Step 3. Click the Sort button in the Sort & Filter group
Z A
A Z

NYC Sort

Step 4. When the Sort dialog box appears:

Select the check box for My data has headers
In the Sort by row:
Select Borough for the Column entry
Select Cell Values for Sort on entry
Select A to Z for Order entry
Click OK
Step 5. Select cells B2:C11 (rows corresponding to the Bronx borough)
Step 6. Click the Insert tab on the Ribbon
Step 7. Click the Insert Scatter (X,Y) or Bubble Chart button in the Charts group
Step 8. When the list chart subtypes appear, click the Scatter button
Step 9. Right-click the chart and choose Select Data…
Step 10. When the Select Data Source dialog box appears:
Select Series1 under Legend Entries (Series)
Click the Edit button Edit

Step 11. When the Edit Series dialog box appears:

Enter Bronx in the box under Series name:
Click OK
Step 12. In the Select Data Source dialog box, click the Add button Add
252 Chapter 6 Exploring Data Visually

Step 13. When the Edit Series dialog box appears

Enter Brooklyn in the box under Series name:
Enter =Data!$B$12:$B$29 in the box under Series X
Enter =Data!$C$12:$C$29 in the box under Series Y values:
Click OK
Step 14. In the Select Data Source dialog box, click the Add button Add

Step 15. When the Edit Series dialog box appears

Enter Manhattan in the box under Series name:
Enter =Data!$B$30:$B$39 in the box under Series X
Enter =Data!$C$30:$C$39 in the box under Series Y values:
Click OK
Step 16. In the Select Data Source dialog box, click the Add button Add

Step 17. When the Edit Series dialog box appears

Enter Queens in the box under Series name:
Enter =Data!$B$40:$B$53 in the box under Series X
Enter =Data!$C$40:$C$53 in the box under Series Y values:
Click OK
Step 18. In the Select Data Source dialog box, click the Add button Add

Step 19. When the Edit Series dialog box appears

Enter Staten Island in the box under Series name:
Enter =Data!$B$54:$B$56 in the box under Series X
Enter =Data!$C$54:$C$56 in the box under Series Y values:
Click OK to close the Edit Series dialog box
Click OK to close the Select Data Source dialog box
Further editing will result in a chart that matches Figure 6.29. Using color, Figure 6.29
distinguishes the 55 observations by their borough. We notice that Manhattan is charac-
terized by two different types of sub-boroughs: one group with low monthly median rent
and moderate percentage of college graduates and one group with high monthly median
rent and high percentage of college graduates. Figure 6.29 suggests that the strength of
the association between monthly median rent and percentage of college graduates may
vary across boroughs. Modifying the coloration of the scatter chart, we can visualize
the strength of the correlation between monthly median rent and percentage of college
graduates for individual boroughs.
For example, suppose we want to highlight the sub-boroughs of Queens on a scatter
chart. In Figure 6.30, we remove the fill color and select a subdued gray color for data
points not corresponding to Queens. We display the correlation between these two vari-
ables for Queens sub-boroughs in the same color (purple) as the Queens data points. This
distinguishes the observations in Queens and emphasizes that, while positive, the strength
of the correlation between monthly median rent and percentage of college graduates is not
as strong in Queens as it is over all boroughs.
While an individual scatter chart is a great way to visualize the relationship between
a single pair of individual relationships, an analyst may want to examine the pairwise
relationships between several different sets of variables. To easily see the relationships
between several pairs of variables at once, a scatter-chart matrix is a visual display of
the scatter charts for every possible pair of quantitative variables in a data set. Figure 6.31
displays the scatter-chart matrix for the NYC data. Each column and row in the scatter-
chart matrix corresponds to one categorical variable. For instance, row 1 and column 1 in
Figure 6.31 correspond to the median monthly rent variable. Row 2 and column 2 corre-
spond to the percentage of college graduates variable. Therefore, the scatter chart shown
in row 1, column 2 shows the relationship between median monthly rent (on the vertical
axis) and the percentage of college graduates (on the horizontal axis) in New York City
sub-boroughs. The scatter chart shown in row 2, column 3 shows the relationship between
the percentage of college graduates (on the vertical axis) and poverty rate (on the horizon-
tal axis).
6-3 Relationships between Variables 253

FIGURE 6.29 Using Color in a Scatter Chart to Communicate a

Categorical Variable

College Graduates (%)

80

70

60

50

40

30

20

10

0
0 500 1000 1500 2000
Median Monthly Rent ($)
Bronx Brooklyn Manhattan Queens Staten Island

FIGURE 6.30 Using Color to Highlight a Specific Value of a Categorical

Variable

Correlation between Rent and Percentage

College Graduates in Queens Weaker than in
Other Sub-boroughs
College Graduates (%)

80
70
60
50
40
30
Correlation = +0.54
20
10
0
0 500 1000 1500 2000
Median Monthly Rent ($)
Bronx Brooklyn Manhattan Queens Staten Island

Figure 6.31 summarizes several interesting findings. Because the points in the scatter
chart in row 1, column 2 generally get higher moving from left to right, this tells us that
sub-boroughs with higher percentages of college graduates appear to have higher median
monthly rents. The scatter chart in row 1, column 3 indicates that sub-boroughs with higher
poverty rates appear to have lower median monthly rents. The data in row 2, column 3
show that sub-boroughs with higher poverty rates tend to have lower percentages of college
graduates. The scatter charts in column 4 show that the relationships between the mean
travel time and the other variables are not as clear as relationships in other columns.
254 Chapter 6 Exploring Data Visually

FIGURE 6.31 Scatter-Chart Matrix for New York City Sub-Borough Data

The scatter charts along the Median Rent Percent College Graduates Poverty Rate Commute Time
diagonal in the scatter-chart
matrix in Figure 6.31 (e.g.,
in row 1, column 1 and in
Median Rent
row 2, column 2) display
the relationship between a
variable and itself. Therefore,
the points in these scatter
charts will always fall along a
straight line at a 45-degree
angle as shown in Figure 6.31.
Percent College Graduates

Excel does not have a Poverty Rate

procedure for generating
a scatter-chart matrix. Each
scatter chart must be created
individually in Excel and
arranged manually.

Commute Time

A table lens is another way to visualize relationships between different pairs of vari-
ables. In a table lens, a table of data is highlighted with horizontal bars with lengths
proportional to the values in each variable’s column. A table lens can be a useful visu-
alization tool for wide and tall data sets as the insight on the relationships between vari-
ables remains evident even if the display is “zoomed out” to show the table in its entirety
or near-entirety.
The following steps demonstrate the construction of a table lens on the quantitative
variables in the New York City sub-borough data.
Step 1. Select cells B2:B56
Step 2. Click the Home tab on the Ribbon
Step 3. Click the Conditional Formatting button in the Styles group
Conditional
Formatting
NYC
From the drop-down menu, select Data Bars and then Blue Data Bar
from the Solid Fill area
Repeat steps 1–3 for the other three variables, changing the cell range in step 1 to C2:C56,
D2:D56, and E2:E56, respectively. This will result in the four variable columns being for-
matted with horizontal bars proportional to the values of the respective variable. To hide
the numeric values in the cells, we execute the following step.
Step 4 hides the values in the Step 4. Select cells B2:E56 then right-click and select Format Cells…
selected cells of the table lens.
Click the Number tab, then select Custom from the Category: area
Reformatting the cell with a
different Category: type will
In the box below Type:, enter ;;;
make the values visible again. Click OK
6-3 Relationships between Variables 255

To visualize relationships between variables with a table lens, we must now sort the data
according to one of the quantitative variables. The next steps sort the data according to
Median Monthly Rent.
Step 1. Select cells A1:F56
Step 2. Click the Data tab on the Ribbon
Step 3. Click the Sort button in the Sort & Filter group
Z A
A Z
Sort

Step 4. When the Sort dialog box appears, select the check box for My data has headers
In the Sort by row:
Select Median Monthly Rent ($) for the Column entry
Select Cell Values for Sort on entry
Select Largest to Smallest for Order entry
Click OK
Figure 6.32 displays the resulting table lens. We interpret this table lens by comparing the
large values to small values pattern in the column we sorted (in this case Median Monthly
Rent) to the patterns in the columns. Because Percentage College Graduates also displays
a large values to small values pattern, we can deduce that this variable has a positive

FIGURE 6.32 Table Lens Using File NYC

256 Chapter 6 Exploring Data Visually

association with Median Monthly Rent. Conversely, Poverty Rate displays a small values
to large values pattern, so we can deduce that this variable has a negative association with
Median Monthly Rent. The Commute Time column displays no pattern, so we can deduce
that this variable has no relationship with Median Monthly Rent. By sorting the table on
the values of a different variable, the relationships between different pairs of variables can
be analyzed.

N otes 1 C omments

1. Excel comes bundled with a Data Analysis add-in that auto- 4. Linear trendlines are appropriate to depict correlation.
mates several statistical procedures. After activating this Linear trendlines on a scatter chart have a constant slope
add-in, Data Analysis can be found in the Analyze group that implies a change in one variable corresponds to a
under the Data tab in the Ribbon. By selecting Data Analy- proportional change in the other variable regardless of
sis and then Correlation, the correlation between each pair the starting values of the variables. In addition to lin-
of quantitative variables can be computed and displayed ear trendlines, Excel supports the fit of several types of
in a matrix. nonlinear trendlines to scatter charts. An exponential
2. Correlation is a measure of association between two quanti- trendline is appropriate when there is a positive (nega-
tative variables. Therefore, it is inappropriate to attempt to fit tive) relationship in which the slope becomes upwardly
a trendline between a quantitative variable and a categorical (downwardly) steeper as you move from left to right on
variable, even if those categories are labeled as numbers the scatter chart. A logarithmic trendline is appropri-
such as 1, 2, 3, … The difference between Category 1 and ate when there is a positive (negative) relationship in
Category 2 may not be the same as the difference between which the slope flattens as you move from left to right
Category 2 and Category 3, so it would be inappropriate on the scatter chart. A polynomial trendline is appro-
to compute a measure such as correlation on these values. priate when the slope changes direction as you move
3. For a set of n observations of two variables (x1, y1); (x2, y2 ); … from left to right on the scatter chart. Figure 6.28 is
(xn, yn) where x– and y– are the sample means for the x and y an example of a polynomial trendline in which the slope
variables, respectively, the formula for computing the sam- changes direction once. There are other polynomial
ple correlation between the two variables is trendlines that allow the slope to change sign multiple
times.

Î (x 1 2 x )(y 1 2 y ) 1 (x 2 2 x )(y 2 2 y ) 1 ... 1 (x n 2 x )( y n 2 y )

n21

Î Î
rxy 5
(x 1 2 x )2 1 (x 2 2 x )2 1 ... 1 (x n 2 x )2 ( y 1 2 y )2 1 ( y 2 2 y )2 1 ... 1 ( y n 2 y )2
n21 n21

6-4 Analysis of Missing Data

Many real-life data sets suffer from missing data values. There are many reasons why data
sets can contain missing values, and it is important to understand the reasons behind the
missing values to know how they should be treated. In this section, we examine some of
the most common forms of missing data, discuss how missing data can be addressed, and
outline how to identify patterns associated with missing data.

Types of Missing Data

Data sets commonly include observations with missing values for one or more variables. In
some cases, missing data naturally occur; these are called legitimately missing data. For
example, respondents to a survey may be asked if they belong to a fraternity or a sorority,
and then in the next question are asked how long they have belonged to a fraternity or a
sorority. If a respondent does not belong to a fraternity or a sorority, the respondent should
skip the ensuing question about how long. Generally, no remedial action is taken for legiti-
mately missing data.
 6-4 Analysis of Missing Data 257

In other cases, missing data occur for different reasons; these are called illegitimately
missing data. These cases can result for a variety of reasons, such as a respondent electing
not to answer a question that the respondent is expected to answer, a respondent dropping
out of a study before its completion, or sensors or other electronic data collection equip-
ment failing during a study. Remedial action is considered for illegitimately missing data.
After detecting illegitimately missing data, the primary options for addressing them are
(1) to discard observations (rows) with any missing values, (2) to fill in missing entries
with estimated values, or (3) treat missing data as a separate category if dealing with a
categorical variable.
Deciding on a strategy for dealing with missing data requires some understanding of why
the data are missing and the potential impact these missing values might have on an analy-
sis. If the tendency for an observation to be missing the value for some variable is entirely
random, then whether data are missing does not depend on either the value of the missing
data or the value of any other variable in the data. In such cases the missing value is called
missing completely at random (MCAR). For example, if a missing value for a question on
a survey is completely unrelated to the value that is missing and is also completely unrelated
to the value of any other question on the survey, the missing value is MCAR.
However, the occurrence of some missing values may not be completely at random. If
the tendency for an observation to be missing a value for some variable is related to the
value of some other variable(s) in the data, the missing value is called missing at ran-
dom (MAR). For data that are MAR, the reason for the missing values may determine its
importance. For example, if the responses to one survey question collected by a specific
employee were lost due to a data entry error, then the treatment of the missing data may be
less critical. However, in a health care study, suppose observations corresponding to patient
visits are missing the results of diagnostic tests whenever the doctor deems the patient too
sick to undergo the procedure. In this case, the absence of a variable measurement actually
provides additional information about the patient’s condition, which may be helpful in
understanding other relationships in the data.
A third category of missing data is missing not at random (MNAR). Data are MNAR
if there is a tendency for a missing entry of a variable to be related to its value. For example,
survey respondents with extremely high or extremely low annual incomes may be less
inclined than respondents with moderate annual incomes to respond to the question on
annual income, and so these missing data for annual income are MNAR.
Whether the missing values are MCAR, MAR, or MNAR, the first course of action
when faced with missing values is to try to determine the actual value that is missing by
examining the source of the data or logically determining the likely value that is miss-
ing. If the missing values cannot be determined, we must determine how to handle them.
Understanding which of these three categories—MCAR, MAR, and MNAR—missing
values fall into is critical in determining how to handle missing data.
If a variable has observations for which the missing values are MCAR, then discarding the
Imputing missing data observations with missing values may be a good choice if there are a relatively small number
using the mode, median, or of observations with missing values. When missing data are MCAR, removing observations
mean changes the resulting with missing values is equivalent to randomly culling the rows of the data set. We will cer-
distribution for that variable.
tainly lose information if the observations that are missing values for the variable are ignored,
In general, the resulting
distribution will have less
but the results of an analysis of the data will not be biased by the missing values. As an alter-
variability than the true native to discarding observations with missing values that are MCAR, it may be useful to
distribution of the variable. replace the missing entries for a variable with the variable’s median, mean, or mode.
If missing data are MAR, then there is a relationship between the likelihood of a vari-
able having a missing value and the value of another variable in the observation. If missing
data are MAR, then discarding observations with missing values can alter the observed
patterns in the remaining data. For example, suppose that missing values of blood pressure
measurements are more likely to occur in younger patients. Then, discarding all records
with missing blood pressure measurements would also discard a higher proportion of
records with younger patients, thereby potentially distorting patterns in the remaining data.
258 Chapter 6 Exploring Data Visually

If missing values for a variable are MAR, it may be possible estimate an observation’s
missing value of the variable based on the values of the other variables in the observation.
If a variable has observations for which the missing values are MNAR, the observation
with missing values cannot be ignored because any analysis that includes the variable with
MNAR values will be biased. Furthermore, there is no satisfactory manner to address a
variable with missing data that are MNAR because it is the (unknown) values of the miss-
ing data that are causing them to be missing. If the variable with MNAR values is thought
to be redundant with another variable in the data for which there are few or no missing
values, removing the MNAR variable from consideration may be an option. In particular, if
the MNAR variable is highly correlated with another variable that is known for a majority
of observations, the loss of information may be minimal.

Exploring Patterns Associated with Missing Data

To demonstrate exploratory data analysis with respect to missing data, we reconsider the
data from Espléndido Jugo y Batido, Inc. As previously noted, there are missing data for the
variables Service Satisfaction Rating and Product Satisfaction Rating in the EJB data. EJB
notes that the entries for the Service Satisfaction Rating and Product Satisfaction Rating are
retrieved from a database that contains customer responses to surveys associated with prod-
uct records. As a customer response to a survey is voluntary, it is reasonable to expect some
missing entries. However, it is important to examine the missing data for these variables to
determine how we should handle them in the data analysis. Specifically, we would like to
gain understanding on whether the missing entries are MCAR, MAR, or MNAR.
In our discussion and demonstrations, we will consider the missing data for the Service
Satisfaction Rating. The only definitive way to determine whether the missing entries
of Service Satisfaction Rating depend on the (unknown) values of Service Satisfaction
Rating or depend on other variables’ values in the same observation would be to obtain
the values of these missing entries. If possible, EJB may want to perform interviews with
customers who did not provide service satisfaction ratings to ascertain that these data are
missing completely at random and were not the result of a customer withholding any sur-
vey response value. For instance, it may be that customers who would otherwise report
middling Service Satisfaction Ratings of 3 are more likely to skip the survey resulting in a
missing entry for this record. However, without obtaining these data, it is difficult to detect
this MNAR pattern of missing data. In many cases, it may not be plausible (or reliable) to
obtain the missing values. In lieu of obtaining the missing values, we use statistical summa-
ries and charts to identify patterns in the missing data for the Service Satisfaction Rating.
We begin by considering the impact of missing data on the distribution of values for
individual variables. To demonstrate, we compare the relative frequency of records by DC
for records in which the Service Satisfaction Rating is reported to the relative frequency
of records by DC for records in which the Service Satisfaction Rating is not reported. We
To clear the content of begin our steps from an empty PivotTable and PivotChart using the Excel Table EJBData
the PivotChart, click the as the source data (as in Figure 6.13).
PivotChart. From the
PivotChart Fields task pane, Step 1. Select any cell in the range of the empty PivotTable, for example, cell A5
clear the contents of the Step 2. When the PivotTable Fields task pane appears, under Choose fields to add
current PivotTable and
to report:
PivotChart by deselecting the
boxes next to the currently
Drag the Order ID field to the Values area
selected field. Drag the Distribution Center field to the Rows area
Drag the Service Satisfaction Rating field to the Filters area
Step 3. Click Sum of Order ID in the Values area
Step 4. Select Value Field Settings… from the list of options, and when the Value
Field Settings dialog box appears:
EJBTable Click the Summarize Values By tab and under Summarize value field
by, select Count
Click the Show Values As tab and in the Show values as drop-down
menu, select % of Grand Total
Click OK
 6-4 Analysis of Missing Data 259

Step 5. In the PivotTable, click the Filter Arrow in cell B1 and in the drop-down
menu:
Select the check box next to Select Multiple Items
Select the check boxes next to 1, 2, 3, 4, and 5
Deselect the check box next to the blank entry (last category)
Click OK
These steps display the frequency distribution of records by DC based on records in which the
Service Satisfaction Rating is reported. This output is displayed in panel (a) of Figure 6.33.
The next steps create the frequency distribution of records by DC based on records in which
the Service Satisfaction Rating is not reported, displayed in panel (b) of Figure 6.33.
Step 6. Select the worksheet containing the previously constructed PivotTable and
PivotChart and make a copy of this worksheet by right-clicking on the work-
sheet name, selecting Move or Copy, and selecting the Create a copy check
box in the More or Copy dialog box
Click OK
Step 7. In the PivotTable, click the Filter Arrow in cell B1 and in the drop-down menu:
Deselect the boxes next to 1, 2, 3, 4, and 5
Select the check box next to the blank entry (last category)
Click OK

FIGURE 6.33 Comparing Frequency of Records by Distribution Center for Records with a
Service Satisfaction Rating in Panel (a) to Frequency of Records by Distribution
Center for Records without a Service Satisfaction Rating in Panel (b)

If the tendency for a record to be missing an entry for Service Satisfaction Rating was
unrelated to the DC from which the record was shipped, we would expect the frequency
distribution in panel (a) of Figure 6.33 to be similar to the frequency distribution in panel
(b) of Figure 6.33. However, we observe that Mississippi was the DC for 8.75% of the
orders with a service satisfaction rating, but Mississippi was the DC for 26.18% of the
orders missing a service satisfaction rating. This finding suggests there may be an associ-
ation between the tendency of a service satisfaction rating to be missing and whether the
260 Chapter 6 Exploring Data Visually

record was shipped from Mississippi. We should be aware that any analysis related to the
service satisfaction ratings for records shipped from Mississippi may be unreliable.
For each variable in the EJB data, we can repeat the process of comparing the frequency
distribution on data for which the Service Satisfaction Rating is reported to the frequency
distribution on data for which the Service Satisfaction Rating is not reported. In these pair-
wise comparisons, we are looking for substantial differences in the distributions that may
provide an indication that the missing Service Satisfaction Rating is related to the value of
the variable being analyzed.
Similarly, by adding Service Satisfaction Rating to the Filters area of a PivotChart, we
can compare the crosstabulation of two or more variables on data for which the Service
Satisfaction Rating is reported to the crosstabulation on data for which the Service Satisfaction
Rating is not reported. In these pairwise comparisons, we are looking for substantial dif-
ferences in the crosstabulations that may provide an indication that the missing Service
Satisfaction Rating is related to variables involved in the crosstabulation.

6-5 Visualizing Time Series Data

Time series data consists of observations collected at different points in time. Contrast this
to cross-sectional data, which consists of observations collected at a single point in time.
In this section, we assume that the observations are collected at regular intervals of time
and ordered chronologically (which is normally the case).
In most cases, time series data should be displayed using line charts. The lines in a line chart
emphasize the sequential flow and relationship between consecutive values, making visible any
trend, variability, and seasonality. Trend is the long-run pattern in values observable over sev-
eral periods of time. Variability in a time series is the difference in values from period to period.
Seasonality is a pattern in time series data that recurs peri­odically over a regular time interval.
In this section, we discuss techniques for visualizing time series data and these characteristics.
An exception to using a line chart for time series data is when we want to emphasize or com-
pare individual values of one or more variables over time. Then a good choice may be to use a
column chart. Figure 6.22 provides an example of a (stacked) column chart in which total annual
sales over time are being compared for different DCs and by customer status (new or existing).

Viewing Data at Different Temporal Frequencies

The Espléndido Jugo y Batido, Inc. (EJB) data are an example of a time series data that we
have already seen in this chapter. Here, we reconsider this data set to consider appropriate
ways to visualize data over consecutive time periods. Suppose that EJB would like to analyze
the temporal pattern of the total sales of pear juice. The following steps create a PivotTable and
PivotChart to examine pear juice sales over the time period of the data in the file EJBTable.
Step 1. Select any cell in the range of the data, for example, cell A3
Step 2. Click the Insert tab on the Ribbon
In the Charts group, click the PivotChart button PivotChart
EJBTable
Step 3. When the Create PivotChart dialog box appears:
Under Choose the data that you want to analyze, choose Select a table
or range and enter EJBData in the Table/Range: box
Under Choose where you want the PivotChart to be placed, select New
Worksheet
Click OK
Step 4. In the PivotChart Fields task pane, under Choose fields to add to report:
Drag the Date Ordered field to the Axis (Categories) area
Drag the $ Sales field to the Values area
Drag the Flavor field to the Filters area
Drag the Category field to the Filters area
Step 5. In the PivotTable, click the Filter Arrow in cell B1 and in the drop-down
menu:
Click Pear
Click OK
 6-5 Visualizing Time Series Data 261

Step 6. In the PivotTable, click the Filter Arrow in cell B2 and in the drop-down
menu:
Click Juices
Click OK
Step 7. With the PivotChart selected, right-click and select Change Chart Type... and
in the Change Chart Type dialog box, select Line
Further editing will result in a chart that matches Figure 6.34. Figure 6.34 displays the total
sales of pear juice on an annual basis. Based on these three years of data, we observe a rel-
atively steady upward trend in sales.
How frequently we plot time series data can dramatically affect what we see. In a
time series chart, the rate at which we display the data (typically along the horizontal axis)
is called the temporal frequency. To demonstrate, we can view the pear juice sales data
at different temporal frequencies by clicking the + – buttons in the bottom right-hand
corner of Figure 6.34. Clicking the plus sign once results in the chart in Figure 6.35.
Figure 6.35 displays the total sales of pear juice on a quarterly basis. As in Figure 6.34,
there is still an upward trend, but now we can see the quarter-to-quarter variability and
observe that quarterly sales have not always increased over time.

FIGURE 6.34 PivotTable and PivotChart of Annual Pear Juice Sales

Time series data in a

PivotTable and PivotChart Clicking the plus sign again results in the chart in Figure 6.36. Figure 6.36 displays the
can be filtered using Excel’s total sales of pear juice on a monthly basis. The jagged lines of Figure 6.36 reveal the
Timeline tool, which functions monthly fluctuations in sales from pear juice.
much like a Slicer. A Timeline
All three line charts of pear juice sales are useful. Aggregating pear juice sales at an
can be added to a PivotChart
by clicking the Timeline
annual level makes longer-term trends more apparent, but it conceals any patterns that may
button in the Filters group on exist at the quarterly and monthly levels. In general, it is good practice to view data at dif-
the Insert tab in the Ribbon. ferent temporal frequencies as you search for interesting patterns.
The aspect ratio is another factor affecting the visualization of line series chart for time
An exception to this series data. As the width of the line chart increases relative to the height of the line chart,
recommendation for the aspect the lines flatten and the variability in the data is suppressed. Conversely, as the height of the
ratio occurs for charts used to
line chart increases relative to the width of the line chart, the lines become spiky and the
show correlation such as scatter
charts, where we recommend
variability in the data is exaggerated. While there is no single aspect ratio that is always
the width to be equal to the the best (as it depends on what you are trying to see), a general suggestion is to make the
height of the chart. chart wider than it is tall.
262 Chapter 6 Exploring Data Visually

FIGURE 6.35 PivotTable and PivotChart of Quarterly Pear Juice Sales

FIGURE 6.36 PivotTable and PivotChart of Monthly Pear Juice Sales

Highlighting Patterns in Time Series Data

To visualize the trend in time series data over time, we can add a trendline. In Figure 6.37, the
linear trendline highlights the upward trend in monthly pear juice sales over time. While such
a trendline can be useful in emphasizing a long-term trend in the data, we should be aware
that the entire range of data is used to fit the trendline. This has particular implications for
time series data, which by nature occurs sequentially over time. In time series data, it is often
desirable to smooth the data in a manner that (1) only uses observations known up to the point
of time at which the smoothing estimate is calculated and (2) does not use observations from
 6-5 Visualizing Time Series Data 263

the distant past. An m-period moving average is computed by averaging the last m values
observed. That is, at a point in time, future observations and observations from more than m
periods in the past are not included in the calculation of the moving average.

FIGURE 6.37 Linear Trendline Indicating Upward Trend in Pear Juice Sales

Total Sales of Pear Juice

Sales ($)
50,000

40,000

30,000

20,000

10,000

Nov
Dec
Oct
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4
2018 2019 2020

The following steps add a three-month moving average trendline to the monthly pear juice chart.
Step 1. Select the data series by clicking on the data points.
Step 2. Right-click and select Add Trendline…
In the Format Trendline task pane, under Trendline Options , select
PearJuiceEJB
Moving Average and enter 3 in the Period box
With additional editing and repeating the preceding steps to create a 6– and 12-month
moving average smoothing, we generate the trellis display of Figure 6.38 showing
monthly pear juice sales with three different moving averages. The top line chart is over-
laid with a three-month moving average. The middle line chart is overlaid with a six-month
moving average. The bottom line chart is overlaid with a 12-month moving average. As
Figure 6.38 shows, as the number of periods on which the moving average is calculated
increases, the more stable the moving average smoothing becomes.
Seasonality is another type of time series pattern of interest in which the values follow a
predictable pattern that repeats itself at regular time intervals. Although the term seasonality
suggests an association with the meteorological seasons, seasonality can be associated with
any regular time interval (hourly, daily, weekly, monthly, quarterly, yearly). For example,
customer arrivals at a restaurant may display a seasonal pattern over a span of hours charac-
terized by spikes at mealtimes and troughs in between. Attendance at amusement parks may
display a seasonal pattern over the week characterized by spikes on the weekend.
Seasonality may be difficult to clearly identify in charts that display all of the data
linearly from oldest to most recent. Instead, the presence of seasonality recurring at a
time interval is often best examined by plotting the data using multiple lines that cor-
respond to a specific time interval. The following steps demonstrate the examination
of monthly seasonality, resulting in the PivotTable and PivotChart in Figure 6.39. We
assume that a PivotTable and PivotChart using the data in the Table named EJBData has
previously been created (corresponding to the analysis in Figure 6.34).
Step 1. Select the worksheet containing the previously constructed PivotTable and
EJBTable PivotChart and make a copy of this worksheet by right-clicking on the
worksheet name, selecting Move or Copy, and selecting the Create a copy
check box in the More or Copy dialog box
264 Chapter 6 Exploring Data Visually

FIGURE 6.38 Trellis Display of Pear Juice Sales Smoothed by Moving

Averages over 3 Months, 6 Months, and 12 Months

Total Sales of Pear Juice: 3-Month Moving Average

Sales ($)
50,000

40,000

30,000

20,000

10,000

0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4
2018 2019 2020

Total Sales of Pear Juice: 6-Month Moving Average

Sales ($)
50,000

40,000

30,000

20,000

10,000

0
Aug

Aug

May

Aug

Nov
Dec
Sep

Sep

Sep
Apr

Oct
Jun
Nov

Nov
Dec

Dec
May

May
Oct

Oct
Jul

Jul

Jul
Apr

Apr
Jun

Jun
Mar

Mar

Mar
Feb

Feb

Feb
Jan

Jan

Jan

Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4
2018 2019 2020

Total Sales of Pear Juice: 12-Month Moving Average

Sales ($)
50,000

40,000

30,000

20,000

10,000

0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec

Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4 Qtr1 Qtr2 Qtr3 Qtr4
2018 2019 2020

Step 2. Click the PivotChart. In the PivotChart Analyze tab on the Ribbon:
Select Field List from the Show/Hide group to activate the PivotChart
Fields task pane
In the PivotChart Fields task pane, clear the contents of the current
­PivotTable and PivotChart by deselecting the boxes next to the currently
selected fields
6-5 Visualizing Time Series Data 265

Step 3. In the PivotChart Fields task pane, under Choose fields to add to report:
Drag the Month Ordered field to the Axis (Categories) area
Drag the Years field to the Legend (Series) area
Drag the $ Sales field to the Values area
Drag the Flavor field to the Filters area
Drag the Category field to the Filters area
Step 4. In the PivotTable, click the Filter Arrow in cell B1 and in the drop-down menu:
Click Pear
Click OK
Step 5. In the PivotTable, click the Filter Arrow in cell B2 and in the drop-down menu:
Click Juices
Click OK
Step 6. With the PivotChart selected, right-click and select Change Chart Type... and
in the Change Chart Type dialog box, select Line
Examining Figure 6.39, there does not appear to be any seasonality on a monthly basis,
which would be observed as a common pattern over months in each of the years.

FIGURE 6.39 Exploring Pear Juice Sales for Seasonality on a Monthly Basis
266 Chapter 6 Exploring Data Visually

Rearranging Data for Visualization

We have considered data sets structured in a manner that every row represents an observa-
tion of a set of variables whose respective values are recorded in columns. In a time series
data set organized in this manner, each row represents an observation of the values of a set
of variables at a specific point in time. However, there are multiple ways to organize the
data within this structure, particularly if the data contain categorical variables.
To demonstrate, consider data on the three-point shot attempts by each team in the
National Basketball Association from 2015 to 2019. Figure 6.40 shows a portion of the same
data arranged in two different ways. In Figure 6.40a, each row consists of three values: an
observation of a team name, a year, and the number of three-point shot attempts by that team
during that year. In Figure 6.40b, each row consists of six values: an observation of a team
name, the number of three-point shot attempts by that team in year 2015, the number of three-
point shot attempts by that team in year 2016, the number of three-point shot attempts by that
team in year 2017, the number of three-point shot attempts by that team in year 2018, and the
number of three-point shot attempts by that team in year 2019.
The arrangement of data in Figure 6.40a is referred to as stacked data. In this panel,
the data are stacked with respect to Year because all values of Year are listed in a single
column. The arrangement of data in Figure 6.40b is referred to as unstacked data. In this
panel, the data are unstacked with respect to Year because there is a column corresponding
to each value of Year.

FIGURE 6.40 A Portion of Data from the File NBA3PA in Stacked (Panel
a) and Unstacked (Panel b) Form

Depending on the source of the data and how it was collected, data may be provided in
stacked or unstacked form. Both arrangements of data can be useful as each arrange-
ment can facilitate different visualizations. For example, the unstacked version of the
facilitates the construction of team-specific line charts. To quickly construct several
line charts to explore interesting patterns in time series data, we construct sparklines.
A sparkline is a minimalist type of line chart directly placed into a spreadsheet cell.
Sparklines are easy to produce and take up little space as they only display the line
 6-5 Visualizing Time Series Data 267

for the data with no axes. The following steps construct sparklines on the data in the
UnstackedData worksheet of the file NBA3PA.
Step 1. Click the Insert tab on the Ribbon
Step 2. Click Line in the Sparklines group
Step 3. When the Create Sparklines dialog box appears:
NBA3PA
Enter B3:F3 in the Data Range: box
Enter G3 in the Location Range: box
Click OK
Step 4. Copy cell G3 to cells G4:G32
The sparklines in column G of Figure 6.41 do not indicate the magnitude of three-point
shot attempts for the various teams, but they do show the overall trend for these data. We
observe that three-point shot attempts appear to be increasing over this five-year period
for almost every team. The Los Angeles Clippers are the only team who appear to have a
decreasing (or at least non-increasing) trend over this period. As can be observed, spar-
klines provide an efficient and simple way to display basic information about a time series.

FIGURE 6.41 Sparklines on the Unstacked Three-Point Shot Attempt Data

268 Chapter 6 Exploring Data Visually

Alternatively, the stacked version of the three-point shot attempt data facilitates the con-
struction of a box and whisker plot in which the Year variable is used as the horizontal
axis. Figure 6.42 displays a box and whisker plot based on the stacked three-point shot
attempt data. A box and whisker plot can be constructed in Excel from unstacked data, but
the different columns are treated as different data series and differentiated by color. Fig-
ure 6.43 displays a box and whisker plot based on the unstacked three-point shot attempt
data. A display like Figure 6.43 may be appropriate for cross-sectional data, but for time
series data, the display in Figure 6.42 is generally preferred.

FIGURE 6.42 Box and Whisker Plot Constructed from Stacked Three-
Point Shot Attempt Data

FIGURE 6.43 Box and Whisker Plot Constructed from Unstacked Three-
Point Shot Attempt Data

NBA Teams Shooting More Three Pointers

Three-Point Attempts
 6-6 Visualizing Geospatial Data 269

Because different arrangements of data can facilitate different visualizations, it is useful to

be able to transform stacked data to unstacked data and vice versa. In the following steps,
we use the file StackedTableNBA3PA to convert the three-point shot attempt data from
stacked form to unstacked form with respect to the Year variable.
Step 1. Select any cell in the range of the Table, for example, cell A1
Step 2. Click the Data tab in the Ribbon
Step 3. Click From Table/Range in the Get & Transform group
StackedTableNBA3PA
Step 4. When the Power Query Editor window appears:
Select the Year column
Click the Transform tab
Click Pivot Column in the Any Column group
When the Pivot Column dialog box appears, select Three Point
Attempts in the drop-down menu for Values Column
Click OK
Click the Home tab and click Close & Load from the Close group
The preceding steps will create an Excel Table in a new worksheet that mirrors the
unstacked data form in Figure 6.40b.
Conversely, in the following steps we use the file UnstackedTableNBA3PA to convert the
three-point shot attempt data from unstacked form to stacked form with respect to the Year
variable.
Step 1. Select any cell in the range of the Table, for example, cell A3
Step 2. Click the Data tab in the Ribbon
Step 3. Click From Table/Range in the Get & Transform group
UnstackedTableNBA3PA
Step 4. When the Power Query Editor window appears:
Select the columns corresponding to 2015, 2016, 2017, 2018, and 2019
(press the Ctrl key while selecting each column)
Click the Transform tab
Click Unpivot Columns in the Any Column group
Click the Home tab and click Close & Load from the Close group
The preceding steps will create an Excel Table in a new worksheet that after sorting on
years, matches the stacked data form in Figure 40a.

6-6 Visualizing Geospatial Data

Geospatial data are data that include information on the geographic location of each
record. When dealing with a geospatial data, displaying them on a map can often help
to discover patterns that would have been otherwise difficult to interpret. In this section,
we consider two general types of geographic visualizations and discuss their use in data
exploration.

Choropleth Maps
Specifically, a choropleth map is a geographic visualization that uses shades of a color,
different colors, or symbols to indicate the values of a quantitative or categorical variable
associated with a geographic region or area. A familiar example of choropleth map is a
weather map, such as the one in Figure 6.44. In Figure 6.44, color is used to depict the
daily high temperature, with warmer colors (on the red end of the spectrum) representing
Choropleth maps are higher temperatures and cooler colors (on the purple end of the spectrum) representing
discussed in Chapters 2 and 9. lower temperatures.
While choropleth maps may provide a good visual display of changes in a variable
between geographic areas, they can also be misleading. If the location data are not granular
enough so that the value of the displayed variable is relatively uniform over the respective
areas over which it is displayed, then the values of the variable within regions and between
regions may be misrepresented. The choropleth map may mask substantial variation of the
270 Chapter 6 Exploring Data Visually

FIGURE 6.44 Weather Map Example

variable within an area of the same color shading. Further, the choropleth map may suggest
abrupt changes in the variable between region boundaries while the actual changes across
boundaries may be more gradual.
Choropleth maps are the most reliable when the variable displayed is relatively con-
stant within the different locations to be colored. If this is not the case and a choropleth
map is desired, the likelihood of the map to convey erroneous insights is mitigated when
(1) variable measures are density based (quantity divided by land area or population) or
(2) the colored regions are roughly equal-sized so there are no regions that are visually
distracting.
For data with variables representing geographical regions (e.g., countries, states, coun-
ties, postal codes), Excel has mapping functionality that will create a choropleth map. For
example, the file IncomeByState contains the median income for each state in the United
Excel’s Filled Map is powered States (Figure 6.45). The following steps use the file IncomeByState to create a choropleth
by the Bing search engine. map using shading to denote the median income in each state.

FIGURE 6.45 A Portion of the Data in the File IncomeByState

Step 1. Select cells A1:B51

Step 2. Click the Insert tab on the Ribbon
Step 3. In the Charts group, click the Maps button Maps
IncomeByState
Step 4. Select Filled Map
With further editing, the preceding steps produce the choropleth map in Figure 6.46.
6-6 Visualizing Geospatial Data 271

FIGURE 6.46 Choropleth Map of United States Displaying Median

Income by Shading States

As Figure 6.46 demonstrates, choropleth maps are typically better for displaying relative
comparisons of magnitude than conveying absolute measures of magnitude. The value of
median income for each state is difficult to estimate from Figure 6.46, but it is relatively
easy to conclude that northeastern states have high median incomes relative to the rest of
the country and many southern states have low median incomes. Indeed, the strength of
a choropleth map is the identification of the high-level characteristics of a variable with
respect to geographic positioning.
We also observe in Figure 6.46 that the relative size of the states plays a role in the audi-
ence’s perception. For instance, Rhode Island is barely visible while larger states dominate
the audience’s field of vision. Another weakness of Figure 6.46 is that it masks the income
distributions within each state.
The file IncomeByCounty contains the median income data at the county level in the
United States (Figure 6.47). The following steps use the file IncomeByCounty to create a
choropleth map using shading to denote the median income in the counties within each
state of the United States.

FIGURE 6.47 A Portion of the Data in the File IncomeByCounty

272 Chapter 6 Exploring Data Visually

Step 1. Select cells A1:C3148

Step 2. Click the Insert tab on the Ribbon
Step 3. In the Charts group, click the Maps button Maps
IncomeByCounty
Step 4. Select Filled Map
With further editing, the preceding steps produce the choropleth map in Figure 6.48.
It may take Excel several
minutes to populate the FIGURE 6.48 Choropleth Map of United States Displaying Median
county-level choropleth map Income by Shading Counties
in Figure 6.48.

The choice of choropleth map (state-level in Figure 6.46 or county-level in Figure 6.48)
depends on the insight intended to be delivered to the audience. These two views could
also be used in tandem by first showing the state-level measure and then selecting one or
more states to “zoom” in on and show the county-level measure.

Cartograms
Another type of geographic display is a cartogram. A cartogram is a map-like diagram
that uses geographic positioning but purposefully represents map regions in a manner that
does not necessarily correspond to land area. For example, Figure 6.49 contains a car-
togram of the United States in which the area of each state is based on its population. A
cartogram often leverages the audience’s familiarity with the geography of the displayed
regions to convey its message. In Figure 6.49, we observe the tiny size of many western
Excel does not have the
states (Alaska, Nevada, Idaho, Nebraska, North Dakota, and South Dakota) and the white
functionality to automatically space paired with the audience’s tacit knowledge of a standard U.S. map conveys the low
generate cartograms. population density in these areas.
A strength of a cartogram is that the area displayed is proportional to the variable being
measured, thus avoiding any misleading impressions. A weakness of a cartogram is that the siz-
ing of the regions according to the displayed variable may distort enough to render the relative
geographic positioning meaningless and the standard area-based geography unrecognizable.
Figure 6.50 contains a specific type of cartogram, known as an equal-area cartogram.
In comparison to the choropleth in Figure 6.46, the equal-area cartogram provides a bal-
anced visual representation of each state while still maintaining fidelity to relative geo-
graphic positioning.
Summary 273

FIGURE 6.49 Cartogram of United States Displaying State Population

FIGURE 6.50 Equal-Area Cartogram of United States Displaying Median

Income by Shading

S U M M A RY

In this chapter, we have discussed the visualization techniques for conducting exploratory
data analysis. We introduced Excel tools for organizing and rearranging data to facilitate
the visual exploration of a data set. We discussed the challenge of missing data. We defined
the types of missing data and the implications on how to potentially address missing
entries.
274 Chapter 6 Exploring Data Visually

We explained and demonstrated the process of exploring data, including the distribu-
tional analysis of individual variables and the crosstabulation of multiple variables. We
visualized the association between pairs of quantitative variables with scatter charts. We
introduced the concept of correlation and described how to use a trendline on a scatter
chart to visualize the sign and strength of the correlation between two variables. We
described how to construct a scatter chart with the consideration of a third (categorical)
variable. To visualize the pairwise association between quantitative variables when
dealing with data sets with several quantitative variables, we introduced the use of a
scatter-chart matrix and the table lens.
In the final two sections, we discussed two particular types of data: time series data and
geospatial data. For time series data, we explained the benefit of line charts to visualize
trend, variability, and seasonality. We presented the use of a moving average to smooth
time series data. For geospatial data, we discussed the strengths and weaknesses of chorop-
leth maps and cartograms.

G L O S S A RY

Aspect ratio The ratio of the width of a chart to the height of a chart.
Cartogram A map-like diagram that uses geographic positioning but purposefully
represents map regions in a manner that does not necessarily correspond to land area.
Categorical variable Data for which categories of like items are identified by labels or
names. Arithmetic operations cannot be performed on categorical variables.
Choropleth map A geographic visualization that uses shades of a color, different colors,
or symbols to indicate the values of a variable associated with a region.
Correlation A standardized measure of linear association between two variables that
takes on values between −1 and +1. Values near −1 indicate a strong negative linear
relationship, values near +1 indicate a strong positive linear relationship, and values near
zero indicate the lack of a linear relationship.
Cross-sectional data Data collected from several entities at the same or approximately the
same point in time.
Crosstabulation A tabular summary of data for two variables. The classes of one variable
are represented by the rows; the classes for the other variable are represented by the columns.
Data cleansing The process of ensuring data is accurate and consistent through the
identification and correction of errors and missing values.
Equal-area cartogram A map-like diagram that uses relative geographic positioning of
regions, but uses shapes of equal (or near-equal) size to represent regions.
Exploratory data analysis (EDA) The process of using summary statistics and
visualization to gain an understanding of the data, including the identification of patterns.
Geospatial data Data that include information on the geographic location of each record.
Illegitimately missing data Missing data that do not occur naturally.
Legitimately missing data Missing data that occur naturally.
Lurking variable A third variable associated with two variables being studied that results
in a correlation between the two variables, falsely implying a causal relationship between
the pair.
Missing at random (MAR) Missing data for which the tendency for an observation to be
missing a value for a variable is related to the value of some other variable(s) in the observation.
Missing completely at random (MCAR) Missing data for which the tendency for an
observation to be missing the value for a variable is entirely random and does not depend
on either the missing value or the value of any other variable in the observation.
Missing not at random (MNAR) Missing data for which the tendency for an observation
to be missing a value of a variable is related to the missing value.
Moving average A method of smoothing time series data that uses the average of the most
recent m values.
Multivariate analysis The examination of patterns by considering two or more variables
at once.
Problems 275

Ordinal variable Data for which categories of like items are identified by labels or names
and there is an inherent rank or order of the categories.
Outlier An unusually small or unusually large data value.
Quantitative scales The range of quantitative values along the horizontal and vertical axes
in a chart.
Quantitative variable Data for which numerical values are used to indicate magnitude,
such as how many or how much. Arithmetic operations such as addition, subtraction, and
multiplication can be performed on a quantitative variable.
Scatter-chart matrix A graphical presentation that uses multiple scatter charts arranged as
a matrix to illustrate the relationships among multiple variables.
Seasonality A pattern in time series data in which the values demonstrate predictable
changes at regular time intervals.
Sparkline A special type of line chart that indicates the trend of data but not magnitude. A
sparkline does not include axes or labels.
Spurious relationship An apparent association between two variables that is not causal,
but is coincidental or caused by the third (lurking) variable.
Stacked data Data organized such that the values for categorical variables are in a single
column.
Table lens A tabular-like visualization in which each column corresponds to a variable and
the magnitude of a variable’s values are represented by horizontal bars.
Tall data A data set with many observations (rows).
Temporal frequency The rate at which time series data is displayed in a chart.
Time series data Data that are collected over a period of time (minutes, hours, days,
months, years, etc.).
Trellis display A vertical or horizontal arrangement of individual charts of the same type,
size, scale, and formatting that differ only by the data they display.
Trend The long-run pattern in a time series observable over several periods of time.
Univariate analysis The examination of the data for an individual variable.
Unstacked data Data organized such that the values for a categorical variable correspond
to labels for separate columns and the columns contain observations corresponding to these
respective category values.
Variability Differences in values of a variable over observations.
Wide data A data set with many variables (columns).

P R O B L E M S

Conceptual
1. Choosing Appropriate Chart. A fitness lab has conducted an experiment in which
100 participants engage in a 30-minute period of high-intensity interval training (HIIT)
and then are asked about their perceived level of exertion. They must select an answer
from 1 to 4, where 1 = “did not feel challenged”; 2 = “broke a sweat but could have
been pushed more”; 3 = “felt challenged but not overwhelmed”; and 4 = “extremely
fatigued.” In addition, each participant’s body fat percentage is recorded. The following
table displays a portion of the data from the experiment. LO 4

Participant Body Fat Percentage Exertion Level

1 27 3
2 31 4
3 24 1
4 21 2
5 14 4
276 Chapter 6 Exploring Data Visually

What is a good way to show how body fat percentage is related to exertion level?
i. Scatter chart depicting relationship between body fat percentage and exertion level
ii. Side-by-side box and whisker plots of the distribution of body fat percentage for
participants at each exertion level
iii. Clustered column chart of body fat percentage with different colored columns for
each exertion level
iv. Sparklines on data unstacked so columns correspond to different exertion levels
2. Performance Evaluation. Kiwi Analytics is assessing two different training programs
for its consulting employees. One group of 50 employees used training method A for
varying numbers of hours and another group of 50 employees used training method B
for varying numbers of hours. Then these employees were evaluated based on their
performance in job-related tasks. The resulting data from this experiment is displayed
in the following two scatter charts. LO 6
Group A Performance
Evaluation Score

1200

1000

800 y = 8.8225x + 213.75

600

400

Correlation = +0.70
200

0
0 10 20 30 40 50
Hours of Training

Group B Performance
Evaluation Score

1200

1000
y = 17.645x + 227.51
800

600

400 Correlation = +0.70

200

0
0 10 20 30 40 50
Hours of Training
Problems 277

Which of the following are accurate statements based on these data? Select all that apply.
i. Training method A and training method B are equally effective at improving
evaluation scores.
ii. Training method A is more effective than training method B because the observa-
tions in the first scatter chart exhibit less variability around the linear trendline than
the observations in the second scatter chart.
iii. Correlation does not imply causation and therefore there is no meaningful conclu-
sion from these charts.
iv. The strength of the linear relationship between hours of training and evaluation
score is the same for both training method A and training method B.
3. Correlation Analysis of Wide Data. Helen Wagner, a marketing analyst for Meredith
Corporation, is examining a data set based on market research of potential patrons of
a new social-media-based e-magazine. The data set has dozens of quantitative vari-
ables measuring characteristics such as patron income, age, daily hours spent on social
media, money spent based on banner ads on websites, etc.
Helen is interested in exploring the association between these variables. What visu-
alization would you recommend? LO 5
i. Scatter-chart matrix
ii. Table lens
iii. Heatmap
iv. Choropleth map
4. Tailored Delivery. Bravman Clothing sells high-end clothing products and is launch-
ing a service in which they use a same-day courier service to deliver purchases that
customers have made by working with one of their personal stylists on the phone. In a
pilot study, Bravman collected 25 observations consisting of the wait time the customer
experienced during the order process, the customer’s purchase amount, the customer’s
age, and the customer’s credit score. The data for these 25 observations were used to
construct the following scatter-chart matrix. LO 6

Wait Time (mins) Purchase Customer Age Credit Score

Amount ($)

Wait Time (mins)

Purchase Amount ($)

Customer Age

Credit Score

Which of the following statements most accurately describes the insight from the scatter-
chart matrix?
i. Purchase amount, customer age, and credit score have pairwise positive relation-
ships with each other.
278 Chapter 6 Exploring Data Visually

ii. Wait time has a nonlinear relationship with purchase amount.

iii. Wait time is negatively related to customer age.
iv. None of the variables appear to be highly correlated with each other.
5. Wage Income Over Lifetime. The Pew Research Center is a nonpartisan think tank
that gathers information on social issues and demographic trends. It has conducted a
longitudinal study in which they collected the wage income of finance professionals
at various points in their lives. After accounting for inflation, the data from this study
were used to create the following chart. LO 5

Wage Income ($1000s)

180

160

140

120

100

80

60

40

20

0
0 20 40 60 80
Age (Years)

What is the best way to visually emphasize the apparent relationship between wage
income and age in this chart?
i. Add a linear trendline to the chart.
ii. Compute the correlation between the two variables and display this on the chart.
iii. Remove the fill color of the data points.
iv. Add a nonlinear trendline to the chart.
6. Wage Income Over Lifetime. Using data from a longitudinal study that collected
wage income of finance professionals at various points in their lives, an analyst has
constructed the following scatter chart and placed a linear trendline on the data. LO 5

Wage Income ($1000s)

180

160

140

120 Correlation = +0.05

100

80

60

40

20

0
0 20 40 60 80
Age (Years)
Problems 279

What is a reasonable conclusion from this chart?

i. There is no relationship between a person’s age and wage income.
ii. A person’s age and wage income do not appear to be linearly related, but there
does appear to be a nonlinear relationship between these two variables.
iii. A scatter chart is not the best way to view these data because they were collected
over time and therefore a line chart should be used.
iv. None of the variables appear to be highly correlated with each other.
7. Types of Missing Data. Label each of the following descriptions based on the type of
missing data being described: missing completely at random (MCAR), missing at ran-
dom (MAR), or missing not at random (MNAR). LO 7
a. A survey is conducted that asks a sample of people their age, highest level of edu-
cation completed, and preference of political party. The results indicate that many
entries for highest level of education completed are missing, and that respondents
were more likely to omit this response if they had a high school education or less.
b. An economic study is conducted asking people to report their profession, job title,
and city of residence. Approximately 10% of the responses for job title are missing,
and it is observed that men are less likely to provide their job title than women.
c. Students at a local college are given a survey to collect data on their study habits
and preferred learning styles. Students complete the survey by filling out a ­Scantron
form on which they fill in circles to specific multiple-choice questions. Due to
Scantron-reading errors, 5% of the question responses are not able to be read
correctly and so are recorded as missing.
d. A clinical study is being done to judge the efficacy of a new drug in reducing a
patient’s high blood pressure. Each month study participants must each undergo a
stress test where they run on a treadmill at maximum effort for 90 seconds, after
which the patient’s blood pressure is recorded. However, each patient is screened
prior to the stress test and if their resting blood pressure is too high, they do not
complete the stress test and the blood pressure reading for that patient is missing.
8. Patterns in Missing Data. Dr. Drake Ramoray is reviewing a patient database and has
noticed that there are some missing data entries. As shown in the table below, he has
highlighted these missing entries as black cells. LO 7

Examine the pattern in this missing data. Which of the following classifications seems
most appropriate?
i. Missing not at random illegitimately
ii. Missing at random
iii. Legitimately missing data
iv. Missing completely at random
9. Smartphone Sales. A manufacturer has collected the sales data for the past four years.
How should the analyst team visualize these data if they are interested in exploring
sales patterns over time? LO 8
i. Pie chart breaking down sales by year
ii. Heatmap to emphasize the times of the year which sales peak
iii. A set of line charts that aggregate data at different time intervals: by year, by
quarter, and by month
iv. Radar chart to represent the cyclical nature of the business year
280 Chapter 6 Exploring Data Visually

10. High-Speed Rail Passengers. Darcy Sears is a manager of a high-speed rail service between
two major metro areas. Darcy is analyzing three years (36 months) of data on the number of
passengers who ride the high-speed train and has created the following chart. LO 8

Based on her experience with the rail service, Darcy believes there is a seasonal pattern
in amount of passenger traffic on the high-speed train. However, Darcy is disappointed
that her chart does not clearly display seasonality. How should Darcy better visualize a
seasonal pattern in the data?
i. Create a different data series for each year consisting of 12 monthly observations
and plot these three separate lines on a new chart with the month labels January
through December on the horizontal axis.
ii. Aggregate the monthly observations into yearly observations and plot the data on
a new chart with Year on the horizontal axis.
iii. Add a linear trendline to the chart above.
iv. Add a moving average with a period equal to the length of the seasonal pattern.
11. Comparing Moving Averages. What is the effect of decreasing the number of periods
in a moving average? LO 8
i. The moving average trendline becomes smoother and less sensitive to fluctua-
tions in the values of the data.
ii. The value of the moving average always decreases.
iii. The moving average trendline becomes more jagged and more responsive to
recent changes in the values of the data.
iv. There is no effect as the moving average depends on the seasonal pattern.
12. Omaha Steaks. As part of a marketing campaign harkening on its company history
based on Nebraska’s beef industry, Omaha Steaks wants to create a visualization dis-
playing the state’s geographic distribution of beef cattle. Using data on the number of
beef cattle in each county in Nebraska, the following choropleth map was created. LO 9
 Problems 281

Which of the following choices best summarizes the strengths and weaknesses of this
visualization?
i. The choropleth map does a good job of depicting the geographic distribution of
beef cattle in the state of Nebraska. However, the color shading does not provide
a distinct enough contrast to identify the counties with the most beef cattle.
ii. The choropleth map does a good job indicating the counties containing the most
cows in absolute terms. However, there is a tendency for the largest counties to
have the most cattle primarily due to their land area, and this can convey a false
differentiation in the density of cows in adjacent counties.
iii. The use of color clearly shows which counties contain the most beef cattle.
However, the large variance in the size of the counties makes it difficult to extract
insight from the visualization for all counties.
iv. All of these are accurate comments.

Applications

13. Tax Data by County. The file TaxData contains information from federal tax returns
filed in 2007 for all counties in the United States (3,142 counties in total). Create an
Excel Table based on these data. Using Excel Table functionality, answer the following
TaxData
questions. LO 1
a. Which county had the largest total adjusted gross income in the state of Texas?
b. Which county had the largest average adjusted gross income in the state of Texas?
14. More Univariate Analysis of EJB Data. In the EJB example discussed within the
chapter, univariate analysis was demonstrated on some of the variables. We continue
this analysis in this problem. LO 2
EJBTable
a. Using a PivotChart, construct the relative frequency distribution of records over the
values of the Category variable. Describe your findings.
b. Using a PivotChart, construct the relative frequency distribution of records over the
values of the New Customer variable. In the PivotTable, relabel a “No” value for
New Customer as “Existing” and a “Yes” value as “New.” Describe your findings.
15. Missing Data for Product Satisfaction Rating. To understand the implications of
missing data, we must explore the patterns associated the missing entries for the Prod-
uct Satisfaction Rating variable in the EJB data used within the chapter. LO 2, 7
EJBTable
a. Construct the relative frequency distribution of records over the values of the Prod-
uct Satisfaction Rating variable. What percentage of records are missing a value of
Product Satisfaction Rating?
b. Considering only records that report values of Product Satisfaction Rating, con-
struct the relative frequency distribution of records over the different flavors.
c. Considering records that are missing values of Product Satisfaction Rating, con-
struct the relative frequency distribution of records over the different flavors.
d. Compare the distributions in part (b) and part (c). What does this comparison
suggest?
16. Business Graduate Salaries. In the file MajorSalary, data have been collected from
111 College of Business graduates on their monthly starting salaries. The graduates
include students majoring in management, finance, accounting, information systems,
MajorSalary
and marketing. LO 2, 3
a. Create a PivotChart to display the number of graduates in each major. Which major
has the largest number of graduates?
b. Create a PivotChart to display the average monthly starting salary for students in
each major. Which major has the highest average starting monthly salary?
17. Federally Insured Bank Failures. The file FDICBankFailures contains data on fail-
ures of federally insured banks between 2000 and 2012. Create a PivotChart to display
a column chart that shows the total number of bank closings in each year from 2000
FDICBankFailures
through 2012 in the state of Florida. Describe the pattern observed in the chart. LO 3
282 Chapter 6 Exploring Data Visually

18. Market Capitalization and Profit. The file Fortune500 contains data for profits and
market capitalizations from a recent sample of firms in the Fortune 500. Prepare a
scatter chart to show the relationship between the variables Market Capitalization and
Fortune500
Profit in which Market Capitalization is on the vertical axis and Profit is on the hori-
zontal axis. Create a trendline for the relationship between Market Capitalization and
Profit. What does the trendline indicate about this relationship? LO 5
19. Market Capitalization and Profit by Sector. The file Fortune500Sector contains data
on the profits, market capitalizations, and industry sector for a recent sample of firms
in the Fortune 500. LO 5
Fortune500Sector
a. Differentiating observations by using a different color for each industry sector, pre-
pare a scatter chart to show the relationship between the variables Market Capital-
ization and Profit in which Market Capitalization is on the vertical axis and Profit is
on the horizontal axis.
b. Emphasize the relationship between Market Capitalization and Profit within the
healthcare sector by formatting all other sectors with data points in gray with no
fill. Create a trendline based only on the observations in the healthcare sector. What
does the trendline indicate about this relationship between Market Capitalization
and Profit within the healthcare sector?
20. Table Lens on Customer Data. Bravman Clothing sells high-end clothing products
and is launching a service in which they use a same-day courier service to deliver pur-
chases that customers have made by working with one of their personal stylists on the
Bravman
phone. In a pilot study, Bravman collected 25 observations consisting of the wait time
the customer experienced during the order process, the customer’s purchase amount,
the customer’s age, and the customer’s credit score. The file Bravman contains the data
for these 25 observations. LO 5
a. Construct a table lens on these data sorted in decreasing order with respect to the
values of the Purchase Amount variable.
b. Summarize the relationships between Purchase Amount and the other three variables.
21. Missing Data in Marketing Survey. The file SurveyResults contains the responses
from a marketing survey: 108 individuals responded to the survey of 10 questions.
Respondents provided answers of 1, 2, 3, 4, or 5 to each question, corresponding to the
SurveyResults
overall satisfaction on 10 different dimensions of quality. However, not all respondents
answered every question. LO 7
a. To highlight the missing data values, shade the empty cells in black.
b. For each question, which respondents did not provide answers? Which question has
the highest nonresponse rate?
22. Smartphone Sales. The file Smartphone contains data on the monthly sales revenue
for a smartphone manufacturer. LO 8
a. Create a line chart to depict the sales time series at the annual level.
Smartphone
b. Create a line chart to depict the sales time series at the quarterly level.
c. Create a line chart to depict the sales time series at the monthly level.
d. What insight do each of these three views provide?
23. Umbrella Sales. The file Umbrella contains data on the quarterly sales revenue for a
manufacturer of umbrellas and other weather-resistant gear. LO 8
a. Create a line chart for the sales time series data. Add a four-period moving average
Umbrella
to the chart to smooth the data.
b. To investigate the possibility of seasonality on a quarterly basis, plot the data as a
collection of five data series (one for each year). From this visualization, do you
observe any indication of seasonality? If so, describe it.
24. Analysis of Scoring in the NFL. The file ScoringNFL contains the points scored by
each National Football League team for 10 seasons. LO 1, 8
a. NFL leadership is interested in seeing how the distribution of scoring across the
ScoringNFL
league has changed over these 10 seasons. Construct a single chart that allows the
 Problems 283

audience to make these comparisons. Provide a thorough discussion of the distribu-

tion of scoring in the league over this decade.
b. NFL leadership is also interested in the temporal patterns in scoring for individual
teams. Unstack the data and use sparklines to show how scoring for each team has
changed over these 10 seasons. List a team with a strong upward trend over these
10 seasons. List a team with a strong downward trend over these 10 seasons.
25. Omaha Steaks (Revisited). In this problem, we revisit the data from Problem 12. As
part of a marketing campaign harkening on its company history based on Nebraska’s
beef industry, Omaha Steaks wants to create a visualization displaying the state’s geo-
NebraskaBeef
graphic distribution of beef cattle. Data on the number of beef cattle in each county in
Nebraska and the size of each county are contained in the file NebraskaBeef. Create a
choropleth map that shades each county in Nebraska relative to the density of beef cat-
tle (as measured by head per square mile). Compare this visualization to the choropleth
map in Problem 12 and describe the strengths and weaknesses. LO 9
Chapter 7
Explaining Visually to Influence with Data
CONTENTS

DATA VISUALIZATION MAKEOVER 7-4 Bringing It All Together: Storytelling

Increasing Youth Participation in Primary Voting and Presentation Design
7-1 Know Your Audience Aristotle’s Rhetorical Triangle
Audience Member Needs Freytag’s Pyramid
Audience Member Analytical Comfort Levels Storyboarding
7-2 Know Your Message Summary
What Helps the Decision Maker? Glossary
Empathizing with Data
Problems
7-3 Storytelling with Charts
Choosing the Correct Chart to Tell Your Story
Using Preattentive Attributes to Tell Your Story

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Explain the importance of knowing your audience’s LO 6 Identify situations for which a slope chart is appropri­
needs and analytical comfort level to create an ate and be able to create a slope chart from data.
effective data visualization or presentation. LO 7 Use preattentive attributes to emphasize certain
LO 2 Explain how to create empathy in the audience insights for an audience and tell a story in a data
with the data to create the most effective visualization.
message possible in your data visualization or LO 8 Define Aristotle’s Rhetorical Triangle and explain
presentation. how it can be employed to connect with the
LO 3 List the types of data visualizations that are most audience to tell a story in a data visualization or
appropriate to communicate specific insights and presentation.
for audiences with different needs and different LO 9 Define Freytag’s Pyramid and explain how this
levels of analytical comfort. provides a suggested structure for telling a story
LO 4 Effectively use dot matrix charts and big-associated in your presentation.
numbers to give the audience a better relative LO 10 Define the concept of storyboarding and explain
understanding of large numerical values in a data how to storyboard a presentation using sticky
visualization. notes and using PowerPoint.
LO 5 Describe examples of how data visualizations can be LO 11 Create an effective story for a presentation that
created to help the audience empathize with data. considers the needs of the audience.
 Data Visualization Makeover 285

D ata V is u ali z ation M a k eover

Increasing Youth Participation in Primary Voting1 1

A 2016 article in The Washington Post discussed In particular, a chart known as a slope chart can
changes in the number of young voters participating communicate this story more effectively. A slope chart
in the U.S. presidential primary elections of 2008 and for these same data is shown in Figure 7.2. The slope
2016. One of the charts included in this article was chart makes it much easier to interpret the changes
similar to what is shown in Figure 7.1. The clustered in youth participation in the primaries from 2008 to
bar chart is used to illustrate that more young voters 2016. The audience can easily see that four of the
participated in the primary election in 2016 than in states (Florida, Illinois, Missouri, and North Carolina)
2008 for four of five states that were closely contested had increases in youth participation, but one state
in the presidential election. (Ohio) had a decrease in youth participation.
We can find several opportunities for improving Figure 7.2 also highlights the difference in the
the design of the chart in Figure 7.1. Based on con­ behavior of youth voters in Ohio versus the other
cepts from previous chapters, we could change some four states using the preattentive attribute of color.
formatting to improve this chart. However, a more In Figure 7.2, we provide further explanation of
substantial criticism of this chart is that it does not the data by using the descriptive title of “Four of Five
effectively explain the data as well as some other Closely Contested States Saw Increased Youth Partic­
chart types. The goal of the chart is to communicate ipation in Primary Elections in 2016.” This explanatory
to the audience that youth participation in primary title greatly reduces the cognitive load on the audience
elections has increased in four of these five states by providing a summary of the insight to be gained from
between 2008 and 2016. It requires considerable this visualization. We have also added some additional
cognitive load for the audience to understand this explanatory information to the chart explaining the
story from the clustered column chart. We can com­ meaning of “youth voters” in this context.
municate this insight more effectively if we choose a Because the main insight to be conveyed in this
different type of chart. chart for The Washington Post article is that most

Figure 7.1 Clustered Column Chart Showing Changes in Youth Participation in United States
Primary Elections from 2008 to 2016
Youth Total Primaries Participation 2008 and 2016

Ohio 425,500
479,400

North Carolina 351,300

221,700

Missouri 221,600
190,900

Illinois 508,200
378,000

Florida 467,400
286,000

0 100,000 200,000 300,000 400,000 500,000 600,000

2016 2008

1
Note that this example is inspired by an example shown here: https://www.crazyegg.com/blog/data-storytelling-5-steps-charts/
(based on data and chart shown here: https://www.washingtonpost.com/news/the-fix/wp/2016/03/17/74-year-old-bernie-sanderss
-amazing-dominance-among-young-voters-in-1-chart/)
286 Chapter 7 Explaining Visually to Influence with Data

states have seen an increase in youth participation in so it is a good choice for these data and audience.
primary elections between 2008 and 2016, Figure 7.2 Using a slope chart for these data emphasizes to the
explains this more effectively to the audience. Slope audience that the important insight being communi­
charts are very effective at showing changes over mul­ cated is the change in primary election participation
tiple variables between two different points in time, between 2008 and 2016.

Figure 7.2 A Slope Chart Showing Changes in Youth Participation

in U.S. Primary Elections from 2008 to 2016
Four of Five Closely Contested States Saw
Increased Youth* Participation in Primary
Elections in 2016
508,200

Ohio, 479,400
467,400

425,500

Illinois, 378,000

351,300

Florida, 286,000

North Carolina, 221,700 221,600

Missouri, 190,900

*
Youth voters are defined as voters who are 30 years old or younger.

2008 2016

Data visualizations are used to explore data or to explain information to the audience.
In this chapter, we focus on data visualizations that are being used to explain data to an
audience. Our goal is to help the audience generate insights from data and to influence the
audience in a way that facilitates better decision making. Explaining data to an audience
through data visualization is similar to the act of storytelling. The best stories take compli-
cated issues, themes, or ideas and convey them to the audience in a manner that is easy for
the audience to understand, captures the interest of the audience, and helps the audience
remember the important issues, themes, or ideas. This is true for stories presented as books,
movies, or even data visualizations. Specific to stories generated from data, storytelling
refers to the ability to build a narrative from the data that is meaningful for the audience,
is memorable for the audience, and is likely to influence the audience.
To be effective at storytelling, we need to understand our audience. We also need to
understand the story, or key insight(s), that we want to convey from the data. Once we
know who our audience is and what story we want to tell, we can then start to think about
7-1 Know Your Audience 287

what type of data visualization is most effective for that audience and that story. We can
also think about specific design attributes and formatting that we should use in the data
visualization to best convey our story to the audience.
In this chapter, we will discuss how we can understand our audience and understand the
story we are trying to convey. We will consider which types of data visualizations are best
for conveying the story. We will introduce several new types of charts, including the slope
chart used in the Data Visualization Makeover for this chapter. We will also revisit several
design issues introduced in previous chapters to illustrate how these design issues can be
used to convey different stories. We conclude this chapter by bringing the concepts together
to discuss broader themes related to storytelling that can be used to help you design effec-
tive presentations with data visualizations.

7-1 Know Your Audience

A data visualization or presentation that is effective for one audience may not be effective
for another audience due to differences in the audience interests, their roles within the
organization, or their level of comfort with analytical methods and tools. Therefore, our
first goal in effective storytelling is to ensure we understand our audience. In particular, we
want to determine (1) the needs of our audience from our data visualization or presenta-
tion; and (2) the level of analytical comfort in our audience.

Audience Member Needs

We should start by understanding the needs of our audience. What does the audience
need to get out of our data visualization or presentation? We will classify these needs
broadly as belonging to one of the following two categories: (1) high-level understanding
or (2) detailed understanding. Audience needs inform which type of story will be most
effective, and hence, suggests certain types of data visualizations may be more effective
than others.
For audiences that only need a high level of understanding from the data, it is best
to use simple charts that clearly communicate the main insight from the data. For audi-
ences that need a more detailed understanding, data visualizations may be more sophis-
ticated so that the audience can understand more details of the analysis.
Consider the case of the Chan Life Insurance Group. Chan Life Insurance sells a
variety of life insurance policies to customers. The company relies on sales agents
to sell its policies. The overall process for completing a sale is (1) the salesperson
requests a quote and sends it to underwriting for pricing; (2) the underwriters perform
a manual evaluation to examine the quote request, evaluate the risk, and appropriately
price the policy; (3) the pricing decision made by underwriting is verified; and (4) the
finalized quote is sent back to the agent who communicates this to the customer.
Chan Life Insurance’s information technology (IT) department is considering invest-
ment in additional technology that would streamline the quote request and underwriting
processes to reduce the delay between when an agent submits a request for a quote and
when the agent receives the finalized quote. The new technology allows for faster quote
request transmission, automates the underwriting process by using built-in algorithms
to evaluate risk and suggest prices, and reduces the amount of time for underwriting
verification.
Chan Life estimates that it will be able to install the new technology system and pro-
vide the required training to all employees by the end of August. Chan Life Insurance
has been collecting data on the amount of time required to respond to requests for
quotes from January through July of the current year. It has also estimated the amount
of time it will take to respond to request for quotes after the new technology is
installed. These data are shown in Table 7.1.
288 Chapter 7 Explaining Visually to Influence with Data

TABLE 7.1  verage Times to Complete Tasks Required for Responding to Request for Quote
A
from Chan Life Insurance Sales Agents
Average Time to Complete Task (minutes)
Task January March May July September* November*
Process Request for Quote 244 230 267 220 70 50
Underwrite and Generate Quote 154 167 172 168 40 20
Verify Quote 98 112 110 115 20 10
Send Quote to Agent 121 115 110 117 120 120
Total Response Time 617 624 659 620 250 200

*Projected

Chan Life Insurance would like to provide updates to its employees on this new pro-
cess, but it must satisfy the needs of several different audiences. One primary audience is
its sales agents. For the sales agents, the most important insight is that the time required
to receive a response for an insurance policy quote request should decrease substantially
after the introduction of the new technology system. The sales agents will need to adjust
their routines so that they can best take advantage of these decreases in response times.
Therefore, the sales agents fall into the category of needing only a high-level understanding
of the data shown in Table 7.1. The most effective chart for this audience is likely to be
something similar to Figure 7.3, which shows a simple line chart displaying the average
total response times by month. Note that here we have differentiated the portions of the line
chart for the months of September and November by using a dashed line rather than a solid
line to highlight the fact that these months represent projected response times after the new
IT system is installed. You can easily change a portion of a line chart by double clicking an
individual data point in the chart to open the Format Data Point task pane, then selecting
the Fill & Line icon and changing the Dash type under Line.

FIGURE 7.3 Line Chart for Chan Life Insurance Sales Agents Audience
Displaying the Total Response Time for Requests for Quotes

New IT System Projected to Decrease Response Time for Quote Requests

Total Response Time (minutes)
700

600 Actual
Projected
500

400

300

200

100

0
January March May July September November
7-1 Know Your Audience 289

Another audience that Chan Life Insurance would like to update is its internal team of
underwriters. The underwriters are likely to want to know more details on how, and why,
the new technology system will reduce the time required to respond to requests for quotes
from sales agents. This new technology will have a direct impact on the work done by
the underwriters, and so they are likely to want a more detailed understanding of the data
shown in Table 7.1. Therefore, a clustered column chart such as the one shown in Figure 7.4
is likely to be more useful to this audience. The chart shown in Figure 7.4 shows the pro-
jected impact of the new IT system on each of the four tasks required to complete a request
for quote from a sales agent. This chart is likely to lead to further dialog and discussion so
that the underwriters can better understand what effect this new system is likely to have on
their work and day-to-day activities.

FIGURE 7.4 Clustered Column Chart for Chan Life Insurance Underwriters
Audience Displaying the Task Times for Responses to Quotes

New IT System Projected to Decrease Processing, Underwriting, and

Verification Times
Average Time to Complete Task (minutes) Process Request for Quote
Underwrite and Generate Quote
300
Verify Quote
Send Quote to Agent
250

200

150

100

50

0
January March May July September* November*
*Projected

Comparing Figure 7.3 to Figure 7.4, we also notice that the chart titles have been
adjusted to provide greater meaning for the intended audience. For the sales agents, Figure
7.3 uses the chart title “New IT System Projected to Decrease Response Time for Quote
Requests” to emphasize that it is important for this audience to recognize that the total time
required to respond to a quote request is expected to decrease substantially after the intro-
duction of the new IT system. Figure 7.4 uses the chart title “New IT System Projected to
Decrease Processing, Underwriting, and Verification Times” to specifically indicate that
the reductions in time occur in the processing, underwriting, and verification steps. Using
descriptive chart titles is an effective way to highlight specific insights for the intended
audience from a data visualization.

Audience Member Analytical Comfort Levels

It is also useful to understand the analytical comfort level of your audience when designing
data visualizations. Some data visualizations can be overly confusing for an audience that
has little experience interpreting more sophisitcated data visualizations. For audiences that
have low comfort with analytics, it is recommended to use simple charts that can be easily
explained and that highlight a single insight. For audiences that have higher levels of com-
fort with analytics, it may be useful to use more sophisticated charts that provide additional
details and insights.
290 Chapter 7 Explaining Visually to Influence with Data

Consider the case of General Hospital, which is located in Olympia, Washington.

General Hospital tracks patient satisfaction scores for all patients that undergo surgery in
its hospital. The patient satisfaction scores are based on a combination of survey responses
from patients on factors such as waiting time, friendliness of staff, and effectiveness of pro-
cedure. The combined scores range between 0 and 5, and they are tracked for each surgical
department in the hospital: General Surgery, Neurologic Surgery, Oncology, Opthalmic
Surgery, Orthopaedic Surgery, Pediatric Surgery, Thoracic Surgery, and Vascular Surgery.
A portion of these data is shown in Figure 7.5.

FIGURE 7.5 Portion of Patient Satisfaction Scores in the File GeneralHospital

GeneralHospital

General Hospital presents regular updates to its staff on the results from these patient
satisfaction surveys. But it must present the results to several different audiences. One of the
audiences has a low analytical comfort level while another audience has a high analytical
comfort level. For the audience with a low analytical comfort level, the column chart shown
in Figure 7.6 is likely most effective in giving a simple overview of the results of the patient
satisfaction scores. Figure 7.6 is easy to interpret and makes it obvious that several surgi-
cal departments (Neurologic, General, and Vascular) have low average patient satisfaction
scores while others (Opthalmic, Thoracic, and Orthopaedic) have high average patient satis-
faction scores. By presenting the data as a sorted column chart, this makes relative compari-
sons among the surgical departments even easier for the audience.
Figure 7.7 shows the same data presented as a box and whisker chart. This type of
data visualization is more appropriate for an audience with a higher analytical comfort
Box and whisker charts are level. Figure 7.7 takes more work for the audience to interpret than Figure 7.6, but it also
discussed in Chapter 5. provides more insights into the data. From Figure 7.7, we see that while the Thoracic and
Orthopaedic Surgical Departments have higher average patient satisfaction scores, the
distributions of these scores are quite different from each other. The patient satisfaction
scores for the Orthopaedic Surgical Department are all quite similar, so there is relatively
little variability in these scores. The patient satisfaction scores for the Thoracic Surgical
Department have much more variability, including several outliers of very low patient sat-
isfaction scores. Figure 7.7 also shows that Ophthalmic Surgical patient satisfaction scores
have little variability while there is considerable variability in the patient satisfaction scores
from Pediatric, Vascular, and General Surgery patients.
7-1 Know Your Audience 291

FIGURE 7.6 Column Chart Displaying Average Patient Satisfaction Scores

for Surgical Patients at General Hospital

Average Patient Satisfaction Scores at General Hospital

Average Score (0 = Very Dissatisfied; 5 = Very Satisfied)
4.5
4.3
4.0
3.7
3.6

2.6
2.4 2.5

Neurologic General Vascular Oncology Pediatric Ophthalmic Thoracic Orthopaedic

Surgical Department

FIGURE 7.7 Box and Whisker Chart Displaying Patient Satisfaction Scores
at General Hospital

Patient Satisfaction Scores by Surgical Department at General Hospital

Neurologic General Vascular Oncology Pediatric Ophthalmic Thoracic Orthopaedic

Average Score (0 = Very Dissatisfied; 5 = Very Satisfied)

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0
292 Chapter 7 Explaining Visually to Influence with Data

While Figure 7.7 conveys quite a bit more information than Figure 7.6, box and whisker
charts can be difficult to interpret for an audience that has a low analytical comfort level.
Therefore, for such audiences we might consider simpler data visualizations, such as
shown in Figure 7.6 using a simple column chart. Alternatively, if we choose to use a
more sophisticated chart such as the box and whisker chart shown in Figure 7.7, we must
remember that we should provide additional details to explain how the audience should
interpret the chart.
Figure 7.8 provides an example of recommendations for the types of charts that may be
most useful for different audiences based on the level of insight needed by the audience
(high level or detailed) and the analytical comfort level of the audience (low or high).
Figure 7.8 shows only general recommendations that will need to be adjusted for the spe-
cific data and audience. Tables and charts can vary substantially within each type based on
the formatting and amount of information contained in the table or chart.

FIGURE 7.8 The Best Data Visualization Depends on the Needs and
Analytical Comfort Level of the Intended Audience

Summary Tables
Violin Charts
Detailed Insights
Clustered & Stacked Crosstabulation Tables
Needed
Bar/Column Charts Geographic Information Box and Whisker
Systems (Maps, Choropleths) Charts
Slope Charts
Strip Charts Scatter Charts
Histograms Sankey Charts
Dot Matrix Charts
Bubble Charts
Simple Bar/Column
Only High-Level Charts Pie Charts Treemaps
Insights Needed
Line Charts

Low Analytical Comfort High Analytical Comfort

7-2 Know Your Message

To best explain your data to an audience, you also need to ensure that you know what
insight(s) you are trying to convey to the audience. This means that you need to understand
the data well enough that you can communicate the insights clearly and succinctly. Not
only should you be able to explain what the data mean to the audience, but you should also
In Chapter 9 we discuss be able to explain the limitations inherent in the data and how these limitations affect the
additional details for how to
insights drawn from the data.
make sure that we are telling
the truth with data and how to
To best explain the data, we need to understand what types of insights will help the
avoid confusing or misleading decision maker. The goal of all analytical methods is to influence the audience in a way
our audience. that facilitates better decisions. Examples of such decisions include how to improve patient
7-2 Know Your Message 293

care in a hospital, how to best communicate with voters in an election, how to allocate
resources among schools in a city to effectively improve educational outcomes, which
types of marketing channels to use to reach the most customers, or where to build a new
store to provide the biggest boost to company sales. To know how to help make our audi-
ence better decisions, we need to understand both our audience and our data well enough
to know what types of insights from the data are most likely to lead to improvements in
decision making.

What Helps the Decision Maker?

To create the most effective data visualization possible, we need to make sure we under-
stand what will help the decision maker. The decision maker is the person (or group
of people) who will have to make decisions based on the analysis presented in a data
visualization. The person designing a data visualization for explanatory purposes is
rarely the final decision maker. The decision maker often must consider several factors
(quantitative and qualitative), assess the impact on a variety of stakeholders, and execute
a decision with incomplete information under time constraints. Therefore, the goal of
an explanatory data visualization is to make insights from the data easy to understand in
a relatively short amount of time so that the decision maker can make the best decision
possible.
Because most decision makers are busy and must make many decisions, it is important
to create data visualizations that are clear and easy to interpret. This is another reason why
it is important to create data visualizations that have high data-ink ratios and are free from
chart junk. However, we also should ensure we understand what types of decisions will be
made based on the analysis of the data visualization.
Let us again consider the case of the Chan Life Insurance Company. As we have
stated, Chan Life Insurance is investing in a new IT system to reduce the amount of time
it takes to respond to a quote request from a sales agent. However, suppose that we learn
that the final choice of which new IT system to install has not yet been made. If we know
that our audience includes the decision maker(s) who must choose among several alterna-
tive suppliers of IT systems, then this helps us understand what type of visualization will
be most effective for this decision. We could consider creating a clustered bar chart sim-
ilar to what is shown in Figure 7.9, which summarizes the performance of the three main
competitors to be the supplier of the new IT system for Chan Life Insurance: YellowHat
Systems, Illumination Software, and Silver Star. Figure 7.9 provides an evaluation of the
three different systems on four criteria: cost, reliability, maintenance support, and ease of
use. This clustered bar chart makes it easy for the decision maker to compare the different
IT software vendors on the different criteria, and it can also lead to further discussion so
that the decision maker can ask for more details on each of the subjective evaluations.
Note that we can display this clustered bar chart in multiple ways. Figure 7.9a groups the
subjective evaluations by IT software vendor. Figure 7.9b groups the subjective evalua-
tions of the different software vendors by criteria to show how vendors compare to each
other on each criterion. Each of these clustered bar charts can be an effective data visu-
alization for the decision maker, or they could be presented together. It is easy to switch
between the two clustered bar chart in Excel. The following steps will change the chart
shown in Figure 7.9a to the chart shown in Figure 7.9b, or vice versa.
Step 1. Right-click the chart and choose Select Data…
Step 2. When the Select Data Source dialog box appears, click the Switch Row/­
Column button Switch Row/Column
294 Chapter 7 Explaining Visually to Influence with Data

FIGURE 7.9 Two Possible Clustered Bar Charts to Assist a Decision Maker
in Choosing among Three Competing New IT Systems for
Chan Life Insurance Company

Comparison of New IT System Providers Ease-of-Use

Software Vendor Maintenance Support
Reliability
Cost
Silver Star

Illumination Software

YellowHat Systems

0 1 2 3 4 5 6 7 8 9 10
Subjective Evaluation (1 = Worst; 10 = Best)
ChanLifeChart
(a)

Comparison of New IT System Providers Silver Star

Evaluation Criterion Illumination Software
YellowHat Systems
Ease-of-Use

Maintenance Support

Reliability

Cost

0 1 2 3 4 5 6 7 8 9 10
Subjective Evaluation (1 = Worst; 10 = Best)
(b)

Empathizing with Data

It is also important to be able to create empathy with data. Empathy refers to the ability
to understand and share in the feelings of others. Being able to empathize with data means
remembering that data are not just numbers from a spreadsheet or database, but that these data
often represent real people, and that the decisions made based on our analysis may have a
substantial impact on real people. Therefore, it is important to consider how we can create data
visualizations that can help others generate empathy. By empathizing with the data, we can
create data visualizations to effectively impact decisions. Two common challenges for creating
empathy with data are that audiences can lose the ability to associate meaning when consider-
ing large numerical values and that it can be difficult to consider individual cases when look-
ing at aggregate statistics. Here, we will discuss ways of dealing with each of these challenges.
Because most of our day-to-day interactions with tangible items are with items that
number less than a few hundred, it can be difficult to fully grasp the relative sizes of large
 7-2 Know Your Message 295

numbers such as values in the millions, billions, or even larger. In general, people can visu-
alize groups of items up to about 100 things. For larger numbers of items, it is difficult for
people to easily visualize the amount.
There are several strategies for helping the audience interpret large numbers. One sug-
gestion is to try to convert the large number into something with which the audience may
be more familiar. For instance, suppose that the next payout from the Powerball Lottery is
projected to be $357 million. This is a large sum of money, and most people cannot easily
imagine its value relative to smaller dollar amounts. However, some simple arithmetic shows
that a payout of $357 million is equivalent to receiving a payout of more than $171,000 per
week for the next 40 years (ignoring inflation). Most people can then compare the $171,000
per week to their own weekly income to get a quick relative calibration for this amount. Or
consider the fact that it is estimated that humans are currently creating approximately 2.5
quintillion bytes of data each day. Most people have no concept of the size of a “quintillion.”
1 quintillion 5 Therefore, it is common to state this as “90% of all the data in existence has been created in
1,000,000,000,000,000,000
the last two years.” This compares the value of interest (the amount of data created each day)
to something to which we can compare its relative size (the total amount of data created over
all time). This can make the value of interest more accessible to the audience.
Another suggestion is to avoid the use of exponential (or scientific) notation for large
numbers in data visualizations. Unless your audience has high analytical comfort and is
familiar with exponential notation, it is best to either use all digits in numerical values or
to use words such as “millions” and “billions” rather than 106 and 109. This is also true for
notation such as 10–3 and 10–6 to indicate values less than 1 because most audiences are not
familiar with easily visualizing these values as 0.00X and 0.00000X.
Many data visualizations such as line charts, bar charts, and column charts also make it
difficult for the audience to comprehend and empathize with the data. Large changes can
be reduced to relatively small changes in the length of a line or height of a column if proper
context is not given to the audience. Consider the case of visualizing the unemployment
rate in the United States. Figure 7.10 shows a simple line chart for the commonly reported
seasonally adjusted unemployment rate in the United States.2 Near the end of the line chart
in Figure 7.10, you can see the effect of the COVID-19 pandemic on unemployment in the
United States. Clearly, the unemployment rate has increased substantially but it can be diffi-
cult for the audience to interpret the magnitude of this change without additional context.

FIGURE 7.10 Simple Line Chart Showing Monthly Unemployment Rate in

the United States Between 1948 and 2020

Monthly Unemployment Rate in the United States, 1948–2020

Percent Unemployed (Seasonally Adjusted)

16

14

12

10

0
20
ly- 8
ly- 1
ly- 4
ly- 7
ly- 0
ly- 3
ly- 6
ly- 9
ly- 2
ly- 5
ly- 8
ly- 1
ly- 4
ly- 7
ly- 0
ly- 3
ly- 6
ly- 9
ly- 2
ly- 5
ly- 8
ly- 1
ly- 4
ly- 7
Ju 194
Ju 195
Ju 195
Ju 195
Ju 196
Ju 196
Ju 196
Ju 196
Ju 197
Ju 197
Ju 197
Ju 198
Ju 198
Ju 198
Ju 199
Ju 199
Ju 199
Ju 199
Ju 200
Ju 200
Ju 200
Ju 201
Ju 201
Ju 201
20
ly-
Ju

2
based on data from U.S. Bureau of Labor Statistics, https://data.bls.gov/
296 Chapter 7 Explaining Visually to Influence with Data

A good way to provide a sense of scale to the audience for large numbers is to use a
dot matrix chart. A dot matrix chart is a simple chart that uses dots (or another simple
graphic) to represent an item or groups of an item. The dots are laid out in a matrix form,
and the size of the matrix is relative to the size of the total number to be conveyed.
Consider the case of the number of jobs lost in the United States in the year 2020 due to the
COVID-19 pandemic. From unemployment data, it is estimated that more than 40 million jobs
were lost in 2020. This is a staggeringly large number. If we were to plot this on a simple line
chart or scatter chart, the value of 40 million would be reduced to a single point on a chart where
the location on the chart corresponds to 40 million. An alternative way to visualize this is to use
a dot matrix chart. We can easily create a dot matrix chart in Excel using the following steps.
We will display jobs in our dot matrix chart using a simple filled dot where will corre-
spond to 100,000 jobs. Because we want to visualize 40 million lost jobs, we will need to
create a matrix of 40,000,000 / 100,000 5 400 dots. We will do this by creating a 20 3 20
matrix of dots. Note that we are not using the Charts function in Excel; instead, we will
simply build this from a blank Excel worksheet.
Step 1. Select cell B3 in a blank worksheet in Excel
Click the Insert tab on the Excel Ribbon
In the Symbols group, click Symbol
Step 2. When the Symbol dialog box opens:
Select Geometric Shapes in the Subset: box and click the filled dot
Click Insert to insert this dot in cell B3
Click Close to exit the Symbol dialog box
Step 3. Copy cell B3 to cells B3:U22
Step 4. Highlight Columns B:U and double-click the edge of one of the columns to
reduce the spacing of the columns
Step 5. Click View in the Excel Ribbon and deselect the check box for Gridlines
Steps 1 through 5 create the dot matrix chart shown in Figure 7.11.

FIGURE 7.11 Initial Dot Matrix Chart Illustrating the Number of Jobs Lost in the United States
During the 2020 COVID-19 Pandemic
 7-2 Know Your Message 297

We can give additional context for these data to the audience by adding additional details
to this dot matrix chart. We can use knowledge of our audience to create a meaningful
reference point for the magnitude of this job loss. Suppose we are aware that our audience
consists of many individuals living in the state of Ohio. According to the U.S. Bureau of
Labor Statistics, the entire labor workforce in the state of Ohio was about 5.8 million jobs in
2020.3 We will incorporate this into our dot matrix chart by using the following steps.
Because each 5 100,000 Step 6. Change the fill color of cells D20:U20 and B21:U22 by clicking the Home
jobs, then 5,800,000 /
100,000 5 58. So, to tab on the Ribbon and then changing the Fill Color to blue
represent 5,800,000 jobs as
(see ­Figure 7.12)
the total workforce in Ohio,
we shade cells D20:U20 and
Step 7. Select cell B1 and enter the text More than 40 Million Jobs Lost in 2020 Due
B21:U22 blue in Step 6, which to COVID-19 Pandemic
corresponds to 58 dots. Select cell B1, click Home on the Ribbon, and change the font to
16 point Bold
Step 8. Select cell D2 and enter the text This is almost 7 times larger than the total
labor force in the State of Ohio using 11 point font
Step 9. Select cell W2 and enter the text 5 100,000 Jobs
Step 10. Select cell V20 and enter the text Total Labor Force in State of Ohio
The shading used in Figure The completed dot matrix chart is shown in Figure 7.12. This chart helps the audience
7.12 is an example of using
get a relative sense of the scope of the job losses by using dots to represent groups of
the Gestalt principle of
enclosure introduced in
100,000 jobs, by using the total labor force in Ohio as a point of comparison, and by giving
Chapter 3. a descriptive chart title. This helps the audience generate empathy with these data.

FIGURE 7.12 Completed Dot Matrix Chart Illustrating the Number of

Jobs Lost in the United States During the 2020 COVID-19
Pandemic Relative to the Total Labor Force in State of Ohio

3
See https://www.bls.gov/eag/eag.oh.htm#eag_oh.f.1.
298 Chapter 7 Explaining Visually to Influence with Data

Another suggestion for helping to create empathy with the data in the audience is to
include a focus on the individual rather than just presenting aggregated statistics. As an exam-
ple, consider that according to the American Society for the Prevention of Cruelty to Animals,
approximately 3.3 million dogs enter animal shelters each year. Of these 3.3 million dogs,
approximately 1.6 million are adopted; 670,000 are euthanized; and 620,000 are returned to
their owners. Figure 7.13 displays these data in the form of a stacked column chart.

FIGURE 7.13 Stacked Column Chart Showing What Happens to the

Approximately 3.3 Million Dogs Who Enter Animal Shelters
Each Year in the United States

Fates of the Approximately 3.3 Million Dogs Who Enter Animal

Shelters Each Year
Number of Dogs (1000s)
3500

Other/Unknown, 410
3000

Returned to Owner, 620

2500

2000 Euthanized, 670

1500

1000
Adopted, 1600
500

Figure 7.13 shows that there are a lot of dogs entering animal shelters each year and that
of those who enter, only some of them are returned to their original owners or adopted by
new families. Clearly, there is a need to try to find more homes for dogs who enter animal
shelters. However, Figure 7.13 may not influence many in the audience to take action if
the audience does not empathize with the data. We can try to create more empathy with
the data by focusing not just on the aggregate statistics but also including something spe-
cific that makes these numbers seem more personable and relatable. One way to do this is
to include pictures in the data visualization. This is even more effective if we can include
individual characteristics so the audience can relate to this specific individual. Figure 7.14
is a revision of Figure 7.13 that includes a picture and additional details.
We can add a picture to an existing chart in Excel by clicking the Insert tab on the

Ribbon, clicking the Pictures button Pictures in the Illustrations group, and then choosing

where we want to take the picture from. The picture used in Figure 7.14 is taken from
Online Pictures
Customer segmentation in marketing provides another example illustrating the impor-
tance of emphasizing the specific to create a story for data. Many companies use market
segmentation analysis to better understand their customers. The basic idea of market seg-
mentation is to divide a company’s customers into different groups that have similar charac-
teristics. Once the company identifies the characteristics that a particular customer group has
7-2 Know Your Message 299

FIGURE 7.14 Revised Stacked Column Chart That Includes a Picture of a

Specific Shelter Dog to Create More Empathy with the Data

Fates of the Dogs Who Enter Animal Shelters Each Year

Number of Dogs (1000s)

3500

Other/Unknown, 410
3000

2500 Returned to Owner, 620

2000 Euthanized, 670 This is shelter dog Max.

Max entered the animal shelter in Portland in

1500 April and he is still available for adoption today.

1000
Adopted, 1600
500

in common, the company can design specific marketing plans, promotions, etc., to appeal
to that specific customer group. A common methodology that companies use to create these
groups is known as clustering. Clustering algorithms use data on the characteristics of cus-
tomers to form groups (or clusters) such that customers within each group share similar char-
acteristics, but customers in separate groups are generally more different from each other.
Consider the case of Third State Bank (TSB). TSB would like to understand its banking
customers better so that it can create better marketing plans that appeal to specific custom-
ers. TSB’s data science group has analyzed thousands of its customers using characteristics
such as age, education level, marital status, number of children, home location (urban, sub-
urban, or rural), and whether or not the customer closely follows sports and politics. TSB
has used a clustering algorithm and found that one of the major clusters for its customers
that is quite different from the other clusters has the characteristics shown in Table 7.2. The
clustering algorithm has arbitrarily assigned this as cluster number 7.

TABLE 7.2 Summary Results from a Clustering Algorithm Used for the
Customers of Third State Bank
Customer Characteristic Cluster #7 Results
Average Age 24
Education Level 67% college or above
Marital Status 72% unmarried
Average Number of Children 0.3
Home Location 67% urban, 21% suburban, 12% rural
Closely Follows Sports? 34% Yes, 66% No
Closely Follows Politics? 53% Yes, 47% No
300 Chapter 7 Explaining Visually to Influence with Data

To create the best story possible for these data, TSB decides to assign the name
“Sophia” to represent the typical customer within this cluster. TSB defines Sophia as a sin-
gle female who lives in an apartment downtown. Sophia is well-educated, follows the news
closely, is likely to be politically active, and does not regularly attend sporting events.
Note that the characteristics assigned to Sophia do not match every customer within
Cluster #7. In fact, these characteristics might not match any of the customers within
Cluster #7. However, it is common for businesses to characterize their different clusters by
giving the customers within the cluster a persona including a name and defining charac-
teristics. The reason for this is that it is much easier to empathize with a character persona
(even when it is fictional) than to empathize with the aggregated customer characteristics
shown in Table 7.2.

7-3 Storytelling with Charts

As we have discussed in this chapter, a goal of storytelling with data is to make it easy
for the audience to interpret the insights from the data and compel the audience to act on
those insights in some way. This starts with understanding the audience and is enabled by
creating empathy with the data. Effective storytelling with data also requires you to use the
correct chart for the data and the insights you are trying to convey to the audience.

Choosing the Correct Chart to Tell Your Story

Consider the case of Opening Horizons Education. Opening Horizons operates a successful
chain of early-enrollment schools and daycares with five locations. The company’s chief
operating officer, Sandy Huggins, is interested in comparing student enrollments at each
location from the start of the current year to the start of the previous year. Sandy has
collected the necessary data for this analysis as shown in Figure 7.15.

FIGURE 7.15 Data Collected to Analyze Student Enrollment at Opening

Horizon Locations

OpeningHorizons

In Sandy’s first attempt to generate a chart to illustrate these data, she created the clus-
tered column chart shown in Figure 7.16. Sandy will present this to the other Opening
Horizons management staff and owners. What she would particularly like to point out to
them is that Durango and Salida enrollments have moved in opposite directions over the
last year while other locations have remained about the same. It is possible to see this in
the clustered column chart in Figure 7.16, but it might not be immediately obvious to the
audience.
 7-3 Storytelling with Charts 301

FIGURE 7.16 Clustered Column Chart Created for the Student Enrollment
Data at Opening Horizons

Comparing Student Enrollments for Opening Horizons Locations

Number of Students at Start of Year

500
Previous Year Current Year

400

300

200

100

0
Durango Gunnison Montrose Pagosa Springs Salida
Location

We introduced a slope chart An alternative chart to display these same data is a slope chart. A slope chart shows the
in the Data Visualization change over time of a single variable for multiple entities by connecting pairs of data points
Makeover at the beginning of
this chapter.
for each entity. Each of the variables that we want to track is plotted on a vertical axis, and
we use the horizontal axis for the dimension over which we want to see the change, or dif-
ference, for each variable. For the Opening Horizons student enrollment data, the variables
are the student enrollments at each location and the dimension is time (previous year and
current year). The change, or difference, for each variable is represented by the slope of the
line in the chart. The steps that follow explain how to create a slope chart for the Opening
Horizons data.
Step 1. Select cells A2:C7
Step 2. Click the Insert tab on the Ribbon
Click the Insert Scatter (X,Y) or Bubble Chart button in the
OpeningHorizons
Charts group
Select Scatter with Straight Lines and Markers
Steps 1-2 produce the chart shown in Figure 7.17.
Step 3. When the default chart appears in Excel, right-click the chart and choose
Select Data…
When the Select Data Source dialog box appears, click the Switch

Row/Column button Switch Row/Column

Click OK
Step 3 changes the chart from what is shown in Figure 7.17 to Figure 7.18.
Step 4. Click anywhere on the chart and click the Chart Elements button
Deselect the check box for Gridlines
Deselect the check box for Legend
Select the check box for Data Labels and choose Right
302 Chapter 7 Explaining Visually to Influence with Data

In Step 5, make sure that you Step 5. Double click the “395” data label for Salida so that only this data label is
double click to select only
selected
the “Previous Year” label; a
single click will select both the
When the Format Data Label task pane appears, click the Label
“Previous Year” and “Current
Year” labels.
Options icon
Click Label Options, select the check box for Series Name under Label
Contains and select Left under Label Position
Step 6. Repeat Step 5 for the “341,” “302,” “198,” and “164” data labels.
Step 7. Click the vertical axis values and click Delete
Step 8. Click the chart title box and change the title to Comparing Student Enrollments
for Opening Horizons Locations
Click the Home tab on the Ribbon
Click the Align Left button in the Alignment group to left justify the
chart title
Change the chart title font to Calibri 16 pt Bold
Drag the chart title box to align with the text to the left of the chart
Step 9. Adjust the size of the chart area and the sizes of the text boxes for the data
labels so that the finished chart appears similar to Figure 7.19
Comparing Figure 7.19 to Figure 7.16 shows that the slope chart makes it much more
obvious to the audience that Durango has experienced an increase in student enrollment
over the last year, Salida has experienced a decrease in student enrollment, and all other
locations have remained relatively constant. This same insight can be seen in Figure 7.16,
but it requires much more cognitive load from the audience to see this in Figure 7.16.
Figure 7.19 makes this story much more obvious to the audience.
The best chart to use in a specific situation depends on the audience, the data, and the
insights to be conveyed. In many cases, different charts can be used to visualize the same
data. It is difficult to define exactly when specific charts will be best for different situations

FIGURE 7.17 Initial Chart Created for the Opening Horizons Location Data

Chart Title
500

450

400

350

300

250

200

150

100

50

0
Durango Gunnison Montrose Pagosa Springs Salida
Previous Year Current Year
7-3 Storytelling with Charts 303

because this depends on things such as the needs of the audience, the analytical comfort
level of the audience, the complexity of the data, and the type of decisions that will be
affected by the insights from the data. In many cases it is best to create several different
draft versions of charts for the same data to judge which chart is most effective given the
needs of the audience and the insights to be explained.

FIGURE 7.18 Switching Row/Column for the Opening Horizons Location

Data Creates a Slope Chart

Chart Title
500

450

400

350

300

250

200

150

100

50

0
Previous Year Current Year
Durango Gunnison Montrose Pagosa Springs Salida

FIGURE 7.19 Completed Slope Chart for the Student Enrollment Data at
Opening Horizons

Comparing Student Enrollments for

Opening Horizons Locations
432
Salida, 395
360
Durango, 341
Pagosa Springs, 302 298

Montrose, 198 202

Gunnison, 164 150

Previous Year Current Year

304 Chapter 7 Explaining Visually to Influence with Data

Using Preattentive Attributes to Tell Your Story

Preattentive attributes refer to the features of a data visualization that can be processed
Preattentive attributes are quickly by the audience. Proper use of preattentive attributes in a data visualization can
covered in detail in Chapter 3. greatly help focus the attention of the audience to explain specific insights and help
craft a story from the data. Preattentive attributes include the attributes of color, form,
spatial positioning, and movement. In this section, we will present several examples for
how preattentive attributes can be used in a data visualization to tell a story about the
data.
Color can be an effective preattentive attribute to use to influence the audience, empha-
size certain insights, or tell a specific story. Figure 7.20 shows quarterly house-price
indexes from 1999–2019 for each state and the District of Columbia. There are a lot of data
represented on this chart, with many possible insights, but it is difficult for the audience
to interpret any particular insight because the chart is complicated. It is difficult to tell any
particular story with the chart shown in Figure 7.20. But we can use color to help empha-
size certain insights and tell a particular story from these data.

FIGURE 7.20 A Line Chart Showing Quarterly House-Price Indexes for

Each State and the District of Columbia from 1999–2019

Quarterly House-Price Index by State AK AL

House-Price Index (Base = 1991) AR AZ
CA CO
600 CT DC
DE FL
GA HI
500 IA ID
IL IN
KS KY
LA MA
400 MD ME
MI MN
MO MS
300 MT NC
ND NE
NH NJ
NM NV
200 OH
HousePricesChart NY
OK OR
PA RI
100 SC SD
TN TX
UT VA
VT WA
0
1999 2004 2009 2014 2019 WI WV
WY
Year

Suppose that we want to emphasize the fact that the state of Florida was impacted more
than most other states by the subprime mortgage crisis in 2008. We can emphasize this
insight in the chart by using color. Figure 7.21 colors the line corresponding to quarterly
house-price indexes in Florida red while using gray for all other states and the District of
Columbia. This makes it easy for the audience to compare the line for Florida to the other
lines. The audience can now see that Florida experienced a large increase in house-price
indexes leading up to 2008 and a large decrease after 2008. We have also changed the chart
title in Figure 7.21 to emphasize this insight and tell this story.
 7-3 Storytelling with Charts 305

We can modify the chart shown in Figure 7.20 to look like the chart in Figure 7.21 by
using the following steps.
Step 1. Click the chart in the file HousePricesChart
Step 2. Double click AK in the chart legend so that only the entry for AK is selected
to open the Format Legend Entry task pane
HousePricesChart
Step 3. When the Format Legend Entry task pane opens
Click the Fill & Line icon
Choose a gray color for Color under Border
Step 4. Repeat Step 3 for all legend entries other than FL
Step 5. Double click FL in the chart legend so that only the entry for FL is selected to
open the Format Legend Entry task pane
Step 6. When the Format Legend Entry task pane opens
Click the Fill & Line icon
Choose Red for Color under Border
Step 7. Click on the chart legend containing the state abbreviations
Press Delete to delete the legend
Step 8. Click the Insert tab on the Ribbon
A
Click Text Box in the Text group and click in the chart just to the
Text
Box
right of the red line for Florida
Type Florida in the text box and change the font to Red A Calibri 10.5
(see Figure 7.21)

FIGURE 7.21 Using Preattentive Attribute of Color to Emphasize the

Effect of Subprime Mortgage Crisis in 2008 on Housing
Prices in Florida

Florida Experienced Extreme Boom and Bust During 2008

Subprime Mortgage Crisis
House-Price Index (Base = 1999)
600

500

400

Florida
300

200

100

0
1999 2004 2009 2014 2019
Year
306 Chapter 7 Explaining Visually to Influence with Data

Size is another useful preattentive attribute that can be used in data visualizations to
explain insights and tell your story. In particular, one simple, but effective, use of size is the
use of a big associated number (BAN) in your data visualization. As the name implies, a
BAN is often associated with BAN is simply a number associated with the visualization that is displayed in very large
a more colloquial phrase, but font. This is an exceptionally simple idea but it can be effective in conveying meaning and
we use a more formal phrase
directing the audience’s attention for telling your story. Recall our example on the number
here that provides similar
meaning.
of dogs that enter animal shelters in the United States in Figure 7.14. We can illustrate the
use of a BAN to emphasize the approximate number of dogs that enter animal shelters
each year as shown in Figure 7.22. In Figure 7.22 we have added “3,300,000” as a BAN to
emphasize this value to the audience. We have also revised the chart title slightly to remove
this redundant information from the title. The preattentive attribute of size for this text in
our data visualization focuses the audience’s attention on this value and then guides the
story to focus on the multitude of dogs that enter shelters each year.
A BAN can be added to an existing chart in Excel by clicking the Insert tab on the
A A
Ribbon and clicking the Text button and selecting Text Box Text to add a text box to
Text Box

the chart. The BAN should then be added using a large font size.

FIGURE 7.22 Chart Incorporating the Use of a BAN to Emphasize the

Number of Dogs that Enter Animal Shelters Each Year in the
United States

Fates of the Dogs Who Enter Animal Shelters Each Year

Number of Dogs (1000s)

3500

Other/Unknown, 410
3000

2500 Returned to Owner, 620

2000 Euthanized, 670 This is shelter dog Max.

Max entered the animal shelter in Portland in

1500 April and he is still available for adoption today.

3,300,000
1000
Adopted, 1600
500
Dogs Enter Animal Shelters Each Year
0 in the United States

7-4 Bringing It All Together: Storytelling

and Presentation Design
The previous sections of this chapter detail how to use a single data visualization to
explain insights to an audience and influence the audience through storytelling. Many
of these same concepts are also applicable to a complete presentation that uses data
visualizations. In this section, we expand on the concepts discussed previously to apply
to full presentations. We will introduce some general concepts for effective presentations,
7-4 Bringing It All Together: Storytelling and Presentation Design 307

and then we will provide some specifics on how to create effective presentations using
Microsoft PowerPoint.
Storytelling has existed for centuries in cultures throughout the world. While the subject
matter of stories differs from culture to culture, there are often common traits that exist
among stories. Storytelling moves beyond simply explaining something to an audience;
a good story draws the audience into it. Good stories are more likely to be remembered by
the audience and are more likely to persuade the audience into taking action.

Aristotle’s Rhetorical Triangle

The famous Greek philosopher Aristotle was one of the earliest to notice that many stories
share similar characteristics. A major requirement of storytelling is being able to connect
with the audience. Aristotle proposed that there are three general areas in which a story
should connect with the audience. Aristotle’s basic ideas are often explained as a Rhetorical
­Triangle as shown in Figure 7.23. The Rhetorical Triangle is a visual illustration defining
the three ways in which a story can connect to the audience: ethos, logos, and pathos.

FIGURE 7.23 The Rhetorical Triangle Describes the Three Ways in Which
an Effective Story Should Connect with an Audience

Ethos

Logos Pathos

Ethos represents the ability to show credibility in a story to your audience. For stories
related to data, this means that your audience trusts that you are presenting the data in a
truthful manner. To be truthful requires that you use accurate data and do not intentionally
hide or distort the insights from that data. For example, ethos can be created through open
disclosure about any data manipulation (for example, treatment of outliers or transforma-
tion of variables). Trust with the audience can be built by clearly explaining the origin of
the data used in the data visualizations and by being open about any assumptions made in
the analysis and presenting alternative interpretations when necessary.
Logos typically refers to the logic or reasoning in the story or presentation. To connect
with your audience using logos, you must present a clear argument on how the data and the
analysis being presented can help the audience make better decisions. It also requires you
to be clear in the logical arguments you are using to move from the underlying data to the
insights that you are emphasizing in your data visualizations and presentation.
The idea of pathos refers to connecting with the audience using emotion. For data
visualizations and presentations related to analytics, you need to generate empathy in your
audience for the data. This can be accomplished by trying to focus on the specific rather
308 Chapter 7 Explaining Visually to Influence with Data

than just the general. Using examples of individuals rather than just talking about aggregate
statistics, using pictures in your data visualizations, and creating context for large numeri-
cal values are all efforts to appeal to the audience using pathos.

Freytag’s Pyramid
The Rhetorical Triangle described by Aristotle reinforces the importance of being able to
connect with the audience. All good storytellers must be able to understand their audience
so that they can connect with them using ethos, logos, and pathos. However, effective sto-
ries often share other characteristics as well. Good stories in many forms including novels,
short stories, plays, movies, and even presentations related to data often follow a similar
recipe to involve the reader, make the insights memorable, and persuade the audience to
take action. This recipe is often summarized using Freytag’s Pyramid, which is shown
visually in Figure 7.24.

FIGURE 7.24 Freytag’s Pyramid Outlines the Common Structure

of Effective Stories

Climax

Rising Action Falling Action

Introduction Conclusion

Freytag’s Pyramid divides a story into five distinct elements: (1) introduction, (2) ris-
ing action, (3) climax, (4) falling action, and (5) conclusion. The introduction presents the
Freytag’s Pyramid was necessary background information so that the audience can understand the story to be told.
developed in 1863 by the This includes explanations of the major characters, the setting, and establishing the basic
German author Gustav
conflict in the story. Rising action begins to outline more details about the major conflict
Freytag.
in the story. It explains the obstacles that are facing the protagonist. The climax is where
the audience is exposed to the major conflict in the story. If the previous elements of the
introduction and rising action have been well executed, the audience should feel involved
enough to care about the outcome to the protagonist during the climax. A good climax has
the audience hoping for a good outcome for the protagonist, but also understanding why
that outcome is not guaranteed. The element of falling action occurs after the climax. The
protagonist’s fate has usually been determined in the climax, and the audience can start to
anticipate how the story will reach its conclusion during the falling action. The conclusion
(sometimes referred to as the denouement or resolution) presents the end of the story. The
conclusion generally resolves the conflicts presented in the story and explains the outcome
for the protagonist.
Many famous works from literature and film follow the structure outlined by Freytag’s
Pyramid including Shakespearean plays, the Harry Potter series of books, and Star Wars
movies. But how does this structure apply to presentations related to data?
 7-4 Bringing It All Together: Storytelling and Presentation Design 309

Consider the case of Hawaiian Bell, a provider of telecommunications services for

residential and business customers in the state of Hawaii. Hawaiian Bell has noticed that
its customer service scores related to internet network outages for residential customers
are substantially lower than industry benchmarks. Hawaiian Bell decided to undertake an
expansive data analytics effort to try to diagnose causes for the lower customer service
scores and to provide recommendations to improve its customer service in this area.
Hawaiian Bell’s data scientists have spent months collecting tens of thousands of records of
data related to customer call logs, customer service requests, and network outages. The group
has spent nearly eight weeks just reconciling the data to remove duplicate entries, resolve
data inaccuracies, and link the data together from multiple sources. Several more weeks were
required for exploratory data analysis to generate descriptive statistics from the data including
the average time for a customer to have connectivity restored, the most common causes of
network outages, the average wait time a customer experiences when reporting an outage, etc.
Hawaiian Bell’s data scientists are now ready to present their results and recommendations.
The audience for this presentation includes several Hawaiian Bell vice presidents who
oversee customer service and network operations. It is also expected that the chief oper-
ating officer, who has final say on any decision, will be in attendance. The data scientists
would like to create an effective presentation for this audience to convince them to take
action to improve Hawaiian Bell’s customer service levels related to residential internet
network outages.
A common mistake made by novice presenters is to present all the analysis that has
been done for a project in the chronological order in which the analysis was completed.
For Hawaiian Bell, that would mean starting with a detailed explanation of the data clean-
ing efforts that were used, then explaining the exploratory data analysis, presenting the
descriptive statistics, and then finally presenting some aggregated findings from these data.
However, this approach is usually not the most effective in terms of telling a story.
Presenting some basic facts A better story to communicate the insights for this project could be similar to the fol-
and explanatory data during lowing. For the introduction, the presenters explain the basics of the problem and why the
the introduction is also an
problem is important. Because all of the audience members are executives at Hawaiian
example of connecting with
the audience using ethos since
Bell, they likely already have a good understanding of the basic operations of the company,
it can help build credibility. but they may not understand much about this specific problem. To help the audience con-
nect with the story, the presenters can introduce a protagonist. In this case, the protagonist
Using a representative could be a fictional residential customer who purchases internet service from Hawaiian
protagonist to which the Bell. The presenters can even create a persona and give this fictional character a name such
audience can relate is an
as Liann. In the introduction, the presenters introduce Liann as a working mother whose
example of using pathos
because it strengthens the
internet is provided by Hawaiian Bell for basic services including email and web access, as
emotional attachment with the well as for accessing online schoolwork for her children. The presenters explain that Liann
audience. represents one of the more than 7,000 residential customers who contacted Hawaiian Bell
in the previous year to report an internet outage.
In the rising action section of the story, the presenters explain that Liann comes home
from work one day to find that there is no internet service available at her home. As the
majority of Hawaiian Bell customers do (68% according to the data analysis), she calls the
Hawaiian Bell “Contact Us” phone number. Liann waits for 12.4 minutes to speak to a live
customer service representative (CSR), which is the average waiting time experienced by
Hawaiian Bell residential customers over the last year. Once she speaks to a live CSR, she
spends 8.9 minutes (the average for all residential customers) explaining her issue, but she
can only give basic information describing the problem (which is true for nearly 84% of
Hawaiian Bell customers who call to report internet outages). Liann is completely unaware
that 16 other Hawaiian Bell customers from her neighborhood have already called to report
network outages at their homes. The presenters explain that this is typically the number of
Hawaiian Bell customers who call to report problems during a typical network outage.
This sets up the climax of our story, which is a conflict between our protagonist, Liann,
and the company in trying to diagnose and resolve an internet outage at Liann’s home. The
presenters explain that Liann hangs up from her call and spends the next 14 hours with no
310 Chapter 7 Explaining Visually to Influence with Data

internet at her home, which was the 80th percentile of internet outage durations experienced
by Hawaiian Bell customers in the previous year. During this time, Liann has no access to
email or the World Wide Web. Her children are not able to access their online assignments
and complete their homework that evening. Liann will call twice more to get updates on
the status of the internet outage, which is actually slightly less than the average number
of times (3.1) that a Hawaiian Bell customer called to get updates on internet outages last
year. After 14 hours of downtime, Hawaiian Bell restores internet access to Liann’s neigh-
borhood, but Liann actually doesn’t know that internet service has been restored until she
returns home from work the next day, 23 hours after her first phone call to Hawaiian Bell.
Liann is also never made aware of the cause of the internet outage, which according to
the data collected is most likely due to a weather-related event (64% of outages) or to a
vehicle-related accident that damages network equipment (17% of outages).
Creating a strong, logical link The climax has provided an immediate resolution to the conflict that was established,
between the problem and but in the subsequent falling action stage, the presenters provide recommendations for
the proposed solution is an
example of using logos to
improving customer service levels. The presenters explain that if Hawaiian Bell had uti-
connect with the audience. lized the incoming information that multiple residential customers in Liann’s neighborhood
had no internet service, then it could provide an automated response that would alert other
customers calling from Liann’s neighborhood that “We have received reports of internet
outages in your area, and we are working on a resolution.” Further, the message could
ask the customers to “Press 2 if you would like to report an internet outage at your home
address.” This would substantially reduce the waiting time for customers because they
would not have to wait to speak to a live CSR, and they would be reassured that Hawaiian
Bell knows about the issue and is working on a resolution. The presenters also explain
that they are recommending that as soon as the cause of the outage is identified and an
estimated time of service restoration has been determined, then the automated message
should be updated to include this information. This would allow Liann to plan accordingly.
Finally, once internet service is restored, the presenters recommend that each customer that
had reported an issue in that neighborhood should receive an automated call and an email
alerting them that internet access has been restored for their home and providing a reason
for the outage. This would inform Liann exactly when service is restored in case she has
any internet-dependent plans and also provides closure for Liann by telling her the cause
for the outage.
The conclusion section of the presentation would outline the costs and benefits of the
recommendations and point out any important assumptions or any known limitations in
the analysis. This provides all the necessary information for the audience to take action
on the recommendations or ask clarifying questions. The conclusion section could also
provide a summary of the fictional ending of the new system for our protagonist, Liann.
The presenters could explain that if these changes would have been in place previously,
Liann would not have wasted nearly 45 minutes of her day calling and waiting to speak
to Hawaiian Bell CSRs. Furthermore, Liann would have been provided much more infor-
mation during the entire episode, which is likely to improve Liann’s view of the customer
service provided by Hawaiian Bell.
Clearly identifying important This story illustrates several important aspects of storytelling. First, it follows the gen-
assumptions and limitations eral outline of the structure from Freytag’s Pyramid. Second, it helps the audience connect
is another example of using
with the story and empathize with the data by focusing not just on the general but also
ethos to connect with the
audience. the specific. Creating the persona of Liann allows the audience to imagine a specific cus-
tomer undergoing specific challenges rather than just presenting aggregated statistics such
as the average customer wait time to report an outage is 12.4 minutes, 68% of customers
experiencing an outage use the “Contact Us” phone number to report internet outages, etc.
Instead, the data-driven analysis is woven into the narrative to provide sufficient details for
the audience to take action or to ask clarifying information.
Note that in this suggested story, we do not explicitly provide all details on the data
cleaning process and exploratory data analysis. This is likely appropriate for this audience,
which is made up of Hawaiian Bell executives. However, if the audience were made up of
7-4 Bringing It All Together: Storytelling and Presentation Design 311

more analytically comfortable analysts who were interested in the details of how we came
to our recommendations, then it could be important to include these details. It is always
important to understand the needs of your audience and their analytical comfort level.

Storyboarding
To create the most effective story for your presentation, it is often useful to develop a
­storyboard. A storyboard is a simple visual organization of the main points of the story
used to provide structure of the narrative that you intend to create for the audience. Story-
boards are commonly used to help develop stories for movies. For presentations related to
data, storyboards help to organize your thoughts and easily move things around to create
the most effective story possible. There are two common methods for creating storyboards:
(1) a low-tech method using sticky notes or (2) a higher-­tech method using a presentation
software such as Microsoft PowerPoint. We will briefly describe both methods for creating
a presentation.
Many storytelling experts strongly recommend using the low-tech method of sticky
notes. Sticky notes are easy to manipulate, do not require a computer, and prevent people
from jumping ahead to designing the final slides for a presentation. To create a storyboard
using sticky notes, all that is required is a pack of sticky notes, pens or pencils for writing,
and a blank workspace. The goal is to provide a visual outline of the main points that you
will communicate to the audience during the presentation. Because sticky notes can be
moved around easily, it is easy to rearrange, add, and delete items from the storyboard.
A partial storyboard for the planned presentation for Hawaiian Bell using sticky notes
could be something like what is shown in Figure 7.25. This storyboard would be used to
craft the overall outline of the final presentation. The presenters can rearrange the sticky
notes, add text to them, remove them, or add new notes to develop the outline. It is impor-
tant that the presenters do not spend time at this stage developing the final visuals for
the presentation. This storyboard is designed to develop the overall structure of the story,
which will often change during this process.

FIGURE 7.25 A Partial Storyboard that Could Be Developed for the

Hawaiian Bell Presentation Using Sticky Notes

INTRODUCE INTRODUCE PROTAGONIST LIANN CALLS

PROBLEM SETTING LIANN CUSTOMER SERVICE
– RESIDENTIAL CUSTOMER – HOME INTERNET – WAITS ON HOLD
SERVICE RATINGS OUTAGE 12.4 MIN
BELOW INDUSTRY – WORKING MOM – NO KNOWLEDGE
STANDARDS W/KIDS OF OTHER
CALLERS

CSR ANSWERS LIANN CALLS BACK

– LITTLE INFO ON TWICE
OUTAGE – INTERNET
– AVERAGE TIME OUTAGE =14 HRS
TALKING TO – NOT NOTIFIED
CUSTOMER = 8.9 MIN WHEN SERVICE
RESTORED
312 Chapter 7 Explaining Visually to Influence with Data

Presentation software such as Microsoft PowerPoint can also be used to develop a

storyboard, but you should be careful not to use the software to build the final slides at this
stage. It is recommended that if you use PowerPoint to build your storyboard, you should
make use of the Slide Sorter view in PowerPoint. This view creates a similar experience to
the manual version of building a storyboard using sticky notes.
An example of creating a partial storyboard for a possible Hawaiian Bell presentation
created using PowerPoint is shown in Figure 7.26. To create this storyboard, each slide is
created using a blank format and the Title and Content layout. We can then view the slides

in Slide Sorter view by clicking the View tab on the Ribbon and select Slide Sorter Slide
Sorter

in the Presentation Views group. Once we are in the Slide Sorter View, we can easily rear-
range slides, delete slides, and add new slides similar to what is done in the manual method
of creating a storyboard using sticky notes.

FIGURE 7.26 A Partial Storyboard that Could Be Developed for the

Hawaiian Bell Presentation Using PowerPoint

N otes 1 C omments

1. For some situations, it may be appropriate to begin your 2. It is possible to add a Storyboarding tool to PowerPoint
presentation with the major recommendations rather than that provides special functionality for building a story-
exactly following the structure outlined in Freytag’s Pyra- board, but this is not a standard part of PowerPoint
mid. This alternative approach is often appropriate when and also requires Visual Basic to be installed on your
the audience includes decision makers who have limited computer.
time and need to know the recommendations immediately
before going into additional details.
 Glossary 313

S u mma r y

In this chapter, we have described methods to help explain data to an audience and to help
the audience to make better decisions. This process starts with understanding the needs
and analytical comfort level of the audience. We discussed how the characteristics of the
audience may affect which types of data visualizations are most effective. We have also
explained the importance of being able to empathize with data to create the most effective
visualizations and presentations. We provided several suggestions to help create empathy
in the audience with the data, including focusing on the specific rather than just the general
and by helping the audience understand large numerical values by giving relative reference
values for comparison. We described how preattentive attributes such as color and size can
be used to highlight particular insights from data and how these can be used in a data visu-
alization to influence the audience. We introduced several new types of data visualizations:
the dot matrix chart and the slope chart.
We also expanded on our discussion of explaining and influencing by connecting these
ideas to storytelling. Being a good storyteller allows you to connect with your audience
and gives you the best chance at influencing the audience to make better decisions. To
illustrate the importance of storytelling and provide suggestions for structuring a presenta-
tion, we introduced the concepts of the Rhetorical Triangle and Freytag’s Pyramid. We also
introduced storyboarding as a key element for developing an effective presentation, and we
showed how storyboards can be created using either sticky notes or using a presentation
software such as PowerPoint.
It is important to realize that every audience and every data set will be unique.
Therefore, each data visualization or presentation will be different to best meet the needs
of the audience and fit what is available in the data. There are many common elements that
generally lead to effective data visualizations, and the information presented in this chapter
attempts to identify those common elements. However, the best method for developing
effective data visualizations and presentations is practice and repetition. The more comfort-
able you become with creating data visualizations and presentations, the more willing you
will be to experiment and to find what works best for you in different scenarios.

G L O S S A RY

Big Associated Number (BAN) A number associated with a visualization that is

displayed in very large font for emphasis or to guide the audience’s attention.
Dot matrix chart A simple chart that uses dots, or another simple graphic, to represent an
item or groups of an item. It is useful for providing additional context to the audience for
large numerical values.
Empathy The ability to understand and share in the feelings of others.
Ethos The ability to show credibility in a story to the audience.
Freytag’s Pyramid Visual illustration that defines the five common elements of
the structure of an effective story: introduction, rising action, climax, falling action,
conclusion.
Logos The ability to connect with the audience through logic and reasoning.
Pathos The ability to connect with the audience using emotion.
Rhetorical triangle Visual illustration proposed by Aristotle to define three general areas
in which a story should connect with an audience: ethos, logos, pathos.
Slope chart A chart that shows the change of a single variable over time for multiple
entities by connecting pairs of data points for each entity
Storyboard A simple visual organization of the main points of the story used to provide
structure of the narrative to be developed for the audience.
Storytelling Specific to stories generated from data, storytelling refers to the ability to
build a narrative from the data that is meaningful for the audience, is memorable for the
audience, and is likely to influence the audience.
314 Chapter 7 Explaining Visually to Influence with Data

P R O B L E M S

CONCEPTUAL
1. Understanding Needs of the Audience. For each of the following audiences, indicate
whether the audience members are more likely to need a high-level understanding or a
detailed understanding from a data visualization. LO 1
a. You are presenting the results of a marketing segmentation study to a group of data
scientists and the company’s Chief Analytics Officer (CAO). The market segmenta-
tion study is designed to determine customer groups for the company to better target
the company’s market to specific customer segments. There are several different al-
gorithms that can be used to perform market segmentation, and the goal of the data
visualization is to get feedback from the data scientists on the different algorithms
and help the CAO make a decision on which algorithm should be used.
b. You are presenting the results of a fundraising analysis to the Board of Directors for
a large nonprofit company. The fundraising analysis is designed to suggest some
additional approaches for increasing the propensity for donors to financially contrib-
ute to the organization. The Board of Directors provides general oversight for the
nonprofit, but it is not involved in day-to-day decision making for the company. The
board is composed of people who are known to be philanthropic and have consider-
able personal resources, but they have little background or expertise in analytics.
c. You are preparing a data visualization that will accompany a public-relations media
release for a project that you recently completed for a startup company that does po-
litical polling and analysis. The project used data collected from likely voters in an
upcoming election, and the goal was to determine the issues that are most important
in terms of influencing who these voters will choose in the next election. The media
release will be sent to a wide range of possible outlets including local television
networks and magazines. The goal of the public-relations release is to generate
publicity for the startup company and make the general public aware of the type of
work this startup company does.
2. Rainfall Amounts by Location. Consider the following two possible data visualizations
that display the results of an analysis of data collected by the meteorology department
of a local university. The data collected are based on the monthly amount of rainfall
received at 10 different locations (labeled as Locations A–J) over the last two years. LO 1

Monthly Rainfall by Location

A B C D E F G H I J
Inches of Rain
6

0
Problems 315

Average Monthly Rainfall by Location

Inches of Rain

4.50 4.51

3.56 3.59
3.27 3.16
2.83

2.04 1.96
1.73

A B C D E F G H I J
Location

a. Which of these two data visualizations (box and whisker chart or column chart)
would be most appropriate for an audience that has a high comfort level with ana-
lytics and would like to understand the variability in monthly rainfall amounts at the
10 locations? Why?
b. Which location has the highest variability in monthly rainfall amounts?
c. Which location has the lowest variability in monthly rainfall amounts?
3. Expense Report Table. The following table shows a monthly expense report summary for
an academic department at a community college located in Bethesda, Maryland. LO 1

Expense Category Expense Amount

Office Supplies $ 32.92
AV & Media Parts/Sup $ 78.90
Computing Expenses $ 348.71
Computer Supplies $ 220.39
Printer Supplies $ 21.55
AV Planning & Consul $ 1,248.50
Video Conferencing $ 29.98
Meeting/Seminar/Event $ 149.33
Refreshments & Meals $ 4,899.03
Travel $ 5,256.30
Adv & Development $ 3,983.75
Marketing/Promotion $ 3,000.00
Postage & Delivery $ 14.88
Memberships & Dues $ 3,600.00
Print/Dupl/Graphics $ 846.47
Parking (NonTravel) $ 861.50
Credit Card Process Fee $ 2,883.51
Services $ 4,136.80
Repair/Maintenance $ 75.00
Electric $ 489.00
316 Chapter 7 Explaining Visually to Influence with Data

Is this data visualization better suited for an audience that needs detailed insights or
high-level insights?
4. Chart to Provide Sense of Scale for Large Numerical Values. Which of the follow-
ing chart types is best used to provide a sense of scale for large numerical values in a
data visualization? LO 4
i. Clustered column chart
ii. Box and whisker chart
iii. Dot matrix chart
iv. Slope chart
5. Chart to Show Changes Over Time of Multiple Entities. Which of the following
chart types can be used to easily show changes to multiple entities over time? LO 3
i. Clustered column chart
ii. Box and whisker chart
iii. Dot matrix chart
iv. Slope chart
6. Ethos, Logos, and Pathos. Match each of the following terms with their correct
explanation. LO 8

Term Explanation
Ethos Connecting to an audience based on logic
Logos Connecting to an audience based on emotion
Pathos Connecting to an audience by establishing credibility

7. Examples of Ethos, Logos, and Pathos. For each of the following examples, indicate
whether it is an example of connecting with an audience using ethos, logos, or pathos.
LO 8
a. Providing references to all raw data used to create your data visualizations so that
the audience understands that the data come from reputable sources.
b. Including pictures of specific types of people affected by the data represented in
your data visualization.
c. Explicitly listing limiting assumptions that were made in creating a particular data
visualization.
d. Using a clear reasoning to connect your data visualization to a recommended action
for the decision maker.
e. Creating context for large numerical values used in a data visualization so the audi-
ence has some relative reference to understand these values.
8. Providing a Storytelling Structure for a Presentation. Which of the following pro-
vides a suggested structure for a story to be used for a presentation? LO 9
i. Freytag’s Pyramid
ii. Rhetorical triangle
iii. Preattentive attributes
iv. Ethos, logos, and pathos
9. Elements of Freytag’s Pyramid. Match each of the following elements from Freytag’s
Pyramid with the correct description of the characteristics of that element. LO 9
Problems 317

Freytag’s Pyramid Element Description

Introduction Highlights the major conflict in the story and presents
the outcome of the conflict
Rising Action Presents the end of the story
Climax Explains the obstacles that are facing the protagonist
Falling Action Connects the resolution of the conflict to how the
story will end
Conclusion Presents the background information for the story and
describes the protagonist

10. BAN and Preattentive Attributes. Using a BAN in a data visualization illustrates the
use of which preattentive attribute? LO 4
i. Size
ii. Color
iii. Shape
iv. Motion
11. Unemployment Rates for New England States. The line chart below shows season-
ally adjusted unemployment rates between 2010 and 2020 for states located in the New
England region of the United States. The designer of this chart wants the audience to
be able to easily compare the unemployment rates during this time frame for Massa-
chusetts to the other states in the New England region. LO 7

Unemployment Rates in New England States

Percent Unemployed (Seasonally Adjusted)
20
18 Connecticut Delaware Maine Massachusetts
New Hampshire Rhode Island Vermont
16
14
12
10
8
6
4
2
0
Nov-10 Apr-12 Aug-13 Dec-14 May-16 Sep-17 Feb-19 Jun-20
Source: https://www.bls.gov/charts/state-employment-and-unemployment/state-unemployment-rates-animated.htm

 ow can the designer of this data visualization use the preattentive attribute of color to
H
make this comparison easier for the audience?
12. Lack of Access to Improved Drinking Water. Improved drinking water sources
are defined as drinking water sources that are protected from outside contamination.
According to the World Health Organization, 663 million people around the world lack
access to sources of improved drinking water, which puts them at high risk for infec-
tion and illness. The bar chart below shows the number of people in each region that
are estimated by the World Health Organization to lack access to improved drinking
water sources. LO 2
318 Chapter 7 Explaining Visually to Influence with Data

Number of People Without Access to Improved Drinking Water Sources

Region
Other

South-eastern Asia

Eastern Asia

Southern Asia

Sub-saharan Africa

0 50 100 150 200 250 300 350

Number of People (millions)
Source: Data from World Health Organization: https://apps.who.int/iris/bitstream/handle/10665/177752
/9789241509145_eng.pdf;jsessionid=D7AEF25BD6C2002814352BD93FCD5607?sequence=1

 hich of the following suggestions would improve this chart in terms of helping the
W
audience empathize with the data? (Select all that are appropriate.)
i. Add a picture showing people who lack access to improved drinking water sources.
ii. Color each bar differently to create more excitement for the audience.
iii. Use a dot matrix chart instead of a bar chart to give the audience a better under-
standing of the magnitudes of these numerical values.
iv. Use a pie chart instead of a bar chart because pie charts create more empathy than
bar charts due to use of preattentive attribute of shape.
v. Use a three-dimensional (3D) bar chart because the depth dimension of a 3D
chart moves an audience to feel more empathy.
13. Sprouts Learning Academy Math Enrichment Program. Sprouts Learning Academy
helps prepare students for standardized tests in math for children in the second- through
fifth-grade levels. Sprouts assesses each child by giving them a pretest when they start
their math enrichment academy program and then again at the conclusion of the acad-
emy program. The test results are measured in percentiles compared to all other students
taking similar tests. The clustered column chart below shows the results of these pre- and
post-academy tests for each grade level. The designer of this chart would like to commu-
nicate to the audience about how well the math enrichment program is doing in improv-
ing the performance of the students on this standardized test at each grade level. LO 6
Effects of Math Enrichment Program at Sprouts Learning Academy
Standardized Test Score Percentile
100
90
Pre-math-enrichment program
80
Post-math-enrichment program
70
60
50
SproutsLearningChart 40
30
20
10
0
2 3 4 5
Grade Level
 Problems 319

Which type of chart could be used to better tell the story about how well the math
enrichment program is doing in improving performance on this standardized test?
i. Pie chart
ii. Slope chart
iii. Dot matrix chart
iv. Box and whisker chart
14. Storyboard Materials. Which of the following are recommended tools for creating a
storyboard to help plan the structure of a presentation? (Select all that apply.) LO 10
i. Microsoft Excel
ii. A calculator
iii. Sticky notes
iv. Microsoft PowerPoint
15. Low-Tech versus High-Tech Storyboards. Which of the following is the foremost
advantage of using a low-tech method versus a high-tech method for creating a story-
board to be used for designing a presentation? LO 10
i. Low-tech methods remove the temptation of being distracted with creating final
slide designs during the storyboarding process so you can concentrate on creating
the structure of the story.
ii. Low-tech methods are much faster to implement than high-tech methods.
iii. Low-tech methods make it easier to turn the finished storyboard into the final
slides for the presentation.
iv. Low-tech methods make it much easier to create the final versions of extremely
sophisticated data visualizations to be used in the final presentation.
16. Storyboarding Goal. Which of the following best describes the goal of the story-
boarding process? LO 10
i. Determine how to use design elements such as preattentive attributes in your data
visualizations to best tell your story.
ii. Create a visual outline of the structure of your presentation.
iii. Put the finishing touches on a presentation by focusing on the format and design
of the final data visualizations.
iv. Provide an opportunity to examine the data to be used to create the data visualiza-
tions to create the most effective charts for your audience.
Applications

17. Unemployment Rates for New England States (revisited). In this problem, we
revisit the chart shown in Problem 11 showing seasonally adjusted unemployment rates
for states in the New England region. Use the preattentive attribute of color to modify
NewEngUnemployChart
this chart so that the audience can easily compare the unemployment rates during this
time frame for Massachusetts to the other states in the New England region. LO 7
18. Lack of Access to Improved Drinking Water (revisited). In this problem, we
revisit the chart shown in Problem 12 showing the number of people without access to
improved drinking water sources by region. LO 4, 7
DrinkingWaterChart
a. Approximately 663 million people do not have access to improved drinking water
sources around the world. Add a BAN to the bar chart in the file DrinkingWaterChart
so that the total number of people in the world who do not have access to improved
drinking water sources is emphasized in this chart.
b. To help create empathy with the data for the audience, create a dot matrix chart to
represent the data in the file DrinkingWaterChart. Use 5 1 million people for
your dot matrix chart. (Hint: To represent 663 million people who lack access to
improved drinking water resources, you will need a matrix that contains 663 .
You can create this by using a 25 3 26 matrix and then adding an additional partial
row of 13 .)
320 Chapter 7 Explaining Visually to Influence with Data

c. Differentiate the colors of the based on the region of the world. Include a legend
on your modified dot matrix chart.
d. The population of the United States is approximately 330 million people. To help give
the audience an even better sense of scale for the number of people around the world
who lack access to improved drinking water, use the Gestalt principle of enclosure by
using the Fill Color function in Excel to shade the cells containing the number of
that correspond to the population of the United States and label this on your chart.
19. Sprouts Learning Academy Math Enrichment Program (revisited). In this ­problem
we revisit the chart shown in Problem 13 for standardized test scores from the pre-math-­
enrichment program and post-math-enrichment program at Sprouts Learning Academy.
SproutsLearningChart
LO 6, 7
a. Create a slope chart for the data in the file SproutsLearningChart that compares
the pre- and post-math-enrichment program standardized test scores for second-
through fifth-grade levels.
b. Test scores have increased for all grade levels between pre- and post-math-enrichment
program tests. However, test scores for fourth-grade level have less difference than
the test scores of second-, third-, and fifth-grade levels. Use the preattentive attribute
of color to emphasize the difference in the results for fourth-grade level compared to
second-, third-, and fifth-grade levels.
20. Engineering Graduate Salaries at Empire State University. Empire State University
(ESU) collects salary data for its engineering college graduates. ESU collects data on
starting salaries for its graduates and then again five years post-graduation. The table
below shows the median starting salaries and median salaries five years post-graduation
by engineering major. Create a slope chart to display these data. LO 6

Salaries
Engineering Major Starting 5 Years Post Graduation
Chemical $72,500 $91,400
Civil $61,000 $70,300
Computer $68,400 $83,600

EngineeringSalaries
Electrical $64,800 $87,400
Industrial $57,400 $74,000
Mechanical $62,900 $77,100

21. Measuring Viral RNA Load. Viral RNA load measures the amount of virus in a speci-
fied volume of bodily fluid. When someone is sick due to a viral infection, the virus rep-
licates rapidly in the body and can be measured based on the viral RNA load. Consider
ViralLoadChart
four possible virus vaccines that are being tested in clinical trials: Vaccines A, B, C, and
D. Each vaccine is given to a group of people and the viral RNA load of each patient is
measured. The study also includes a Control group of patients who receive only a pla-
cebo vaccine. The epidemiologists running these vaccine trials would like to present the
results of the studies using a box and whisker chart for an audience of clinicians who are
familiar with how to interpret box and whisker charts. The epidemiologists would like to
emphasize the results for Vaccine C compared to the results for the Control group. Mod-
ify the box and whisker chart found in the file ViralLoadChart using the preattentive
attribute of color to emphasize the results of Vaccine C and the Control group to make it
easier for the audience to make the comparison between these data. LO 7
22. Defining Freytag’s Pyramid Structure for a Story. Think of one of your favorite
books, movies, or plays. Divide up the story that is told in this book, movie, or play
according to the elements outlined in Freytag’s Pyramid: Introduction, Rising Action,
Climax, Falling Action, Conclusion. For each element briefly describe what part of the
story is included in this element and how it fits into the defined element. LO 9
Problems 321

23. Storyboard for Reducing Waiting Time at General Hospital. The Division of Per-
formance Improvement and Analytics at General Hospital has been working on a four-
month project to provide recommendations for reducing the waiting time experienced
by patients who arrive to its Pediatrics Health Care Wing that provides for presched-
uled clinical services such as checkups and wellness visits. The project began with an
intensive six-week data collection effort that measured the waiting time for patients
as well as the patient satisfaction with their visit and any required follow-up visits.
The next four weeks were spent cleaning the data and matching the data with existing
patient and healthcare provider data from the hospital’s IT system. The remainder of
the time was spent analyzing the data, discussing findings with the clinicians, and for-
mulating final recommendations.
The team is now ready to present its findings and recommendations to the senior
leadership of the hospital, including the hospital’s Chief Operating Officer (COO)
who has final decision-making power to decide which, if any, recommendations to be
followed. Some of the interesting findings identified by the team from the Division of
Performance Improvement and Analytics include the following:
●● Patients who were scheduled in the first or second appointment of the day experienced

an average waiting time of 5.8 minutes. Patients who were scheduled for the next-to-
last or last appointment of the day experienced an average waiting time of 42.3 minutes.
●● Patients who responded to a reminder call about their appointment had a no-show

rate (meaning they did not show up at the scheduled time of their appointment) of
8.5%. Patients who did not respond to a reminder call about their appointment had a
no-show rate of 19.3%.
●● Patient no-show rates were slightly higher in the afternoon than in the morning and

higher on Fridays than any other weekday.

●● Seventy-four percent of patients who received a text reminder responded to the reminder

while only 51% of patients who received a phone call responded to the reminder.
●● When patients did not show up for their scheduled appointment time, patients who

did not have a prescheduled appointment were often given priority to fill this slot.
However, these patients required an average appointment duration that was 1.7 times
as long as prescheduled patients because they had to complete additional tasks such
as filling out medical history forms.
●● There was substantial variability in the waiting times experienced by patients across

different clinicians. Patients waiting to see Dr. Martinez had the highest average
waiting times, which were 1.5 times as long as the average waiting times for
Dr. Ahuja, whose patients had the shortest average waiting times.
●● Patient waiting times are highly correlated with their satisfaction-survey scores.

Patients who waited longer had substantially lower patient-satisfaction scores.

●● Patients seen in the morning were 17% more likely to have a follow-up visit sched-

uled than those seen in the afternoon.

Based on these findings, the team would like to make the following recommendations:
●● All patients should receive a text message in addition to a phone call about an up-

coming appointment.
●● A new team should be created to examine the creation of standards for pediatric

patient checkups and wellness visits to (1) reduce the variability experienced by
patients waiting to see different clinicians; (2) examine the cause of why patients
seen in the morning are scheduled for follow-up visits more often than patients seen
in the afternoon.
●● Allow more time between scheduled patient visits in the afternoon than in the morn-

ing to prevent long waits for patients arriving in the afternoon.

Create a storyboard to outline the structure of the presentation for this audience of
senior hospital executives including the COO. Make sure that your storyboard clearly
defines the protagonist in your story and provides sufficient detail that it could be used
to create a final presentation. LO 10, 11
Chapter 8
Data Dashboards
CONTENTS

DATA VISuALIzATION MAkEOVER 8-4 USING EXCEL TOOLS TO BUILD A DATA

Washington State Transportation DASHBOARD
Improvement Board Espléndido Jugo y Batido, Inc.
Using PivotTables, PivotCharts, and Slicers to
8-1 WHAT IS A DATA DASHBOARD?
Build a Data Dashboard
Principles of Effective Data Dashboards
Linking Slicers to Multiple PivotTables
Applications of Data Dashboards
Protecting a Data Dashboard
8-2 DATA DASHBOARDS TAXONOMIES Final Review of a Data Dashboard
Data Updates
8-5 COMMON MISTAKES IN DATA DASHBOARD
User Interaction
DESIGN
Organizational Function
8-3 DATA DASHBOARD DESIGN SUMMARY
Understanding the Purpose of the Data GLOSSARY
Dashboard PROBLEMS
Considering the Needs of the Data Dashboard’s
Users
Data Dashboard Engineering

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Explain what a data dashboard is. LO 5 Describe and explain the principles of data dash-
board design and development.
LO 2 Describe and explain the principles of effective
data dashboards. LO 6 Use Excel tools to build a data dashboard.
LO 3 List common areas of application of data LO 7 List common mistakes made in data dashboard
dashboards. design and development.
LO 4 Describe and explain various data dashboard
taxonomies.
 Data Visualization Makeover 323

D ata V is u ali z ation M a k eover

Washington State Transportation Improvement Board

The Washington State Transportation Improvement This dashboard has many positive features. The
Board (TIB) is an independent state agency responsi- information provided on the TIB At A Glance page of
ble for distributing and managing street construction this dashboard is easy to read. It appears to use space
and maintenance grants throughout Washington effectively; it is not crowded and does not appear
State. TIB selects, funds, and administers transporta- cramped. It also uses tabs to allow for the presenta-
tion projects that best address the criteria established tion of different measures related to the operation of
by the Board. the TIB on separate screens. However, there is room
TIB has created and maintains the Transportation for improvement.
Improvement Board Performance Management Dash- Consider the information in the pane on the left
board to provide the public with up-to-date informa- side of this data visualization. This information may
tion on the status of its various projects. A portion of be useful, but it lacks the context necessary to under-
the TIB At A Glance page of this dashboard is pro- stand and interpret it. Are these year-to-date values?
vided in Figure 8.1. If so, is the basis a calendar year or a fiscal year? If the
This dashboard shows breakdowns of the Finan- basis is a fiscal year, how is the fiscal year defined?
cial Status, Project Status, and KPI (key performance The numbers are also provided in rounded rectangles
indicator) Status in the pane on the left side of the that are filled with red, green, gold, or white without
dashboard. A county map of the state is provided in a readily available indication of what these colors rep-
the pane on the right side of the dashboard. Directly resent. Furthermore, the use of red and green as prin-
above the map, there are several tabs that corre- cipal colors makes it difficult for colorblind members
spond to various factors related to the operation of of the audience to visually process this portion of the
TIB (Inventory, Fund Balances, Gas Tax Revenues, chart, although the dashboard’s inclusion of the actual
Accounts Payable, and Commitment). Note that we numbers would mitigate this problem if context were
are currently on the Inventory tab; in this discussion, provided.
we will focus on the data visualization provided on Now consider the map in the pane on the right
this tab as shown in Figure 8.1. side of the dashboard. Each county is red, green, or

Figure 8.1 The Tib At A Glance Page of the Transportation Improvement Board Performance Management Dashboard

Financial Status Inventory Fund Balances Gas Tax Revenue Accounts Payable Commitment
TIA Fund Balance
SCPSA Fund Balance Current Inventory Status
Project Payments (MTD)
Remaining Commitment
Change in TIB Funds (MTD)

Project Status
Ac ve Projects
Completed Projects (FY)
Under Construc on

KPI Status
TIB Commitment Level
Net Revenue
Construc on Payments
Average Payment Cycle
Delayed Projects
Transac on Processing

Rollover map for details

(Continued)
324 Chapter 8 Data Dashboards

gray. Again, the dashboard provides no immediate We can address these problems by making a few
indication of what these colors represent, and the use relatively minor modifications to this data visualization.
of red and green as principal colors makes it difficult By adding a title to the left pane, we provide context
for colorblind members of the audience to visually for the information provided in that pane. By using
process this portion of the chart. Furthermore, two of different color schemes in the two panes and adding
these colors (red and green) are identical to two colors legends to explain the uses of these colors, we elim-
used in the pane on the right side of this data visual- inate the risk of confusion that exists in the original
ization. Does this convey specific information or is it an dashboard. By avoiding the use of both red and green
arbitrary choice? in either panel, we also make the dashboard easier
On the positive side, we can open a popup box with for colorblind audience members to interpret. And by
the name of a county and more details on the TIB’s adding the county names to the map in the right pane,
activity in that county by rolling the cursor over the we make it easier for users to identify and find infor-
county on the map. However, this is the only way we can mation on specific counties. These changes, as shown
find the names of the individual counties on this map. in Figure 8.2, improve this dashboard substantially.

Figure 8.2 Improved TIB At A Glance Page of the Transportation Improvement Board Performance Management Dashboard

State Totals (calendar year to date) Inventory Fund Balances Gas Tax Revenue Accounts Payable Commitment

Financial Status Current Inventory Status

TIA Fund Balance
Pend
SCPSA Fund Balance Whatcom
Oreille
Project Payments (MTD) Okanogan
Skagit Ferry
Remaining Commitment
Change in TIB Funds (MTD)
Clallam Snohomish Stevens
No At Risk Projects
Project Status Chelan
Acˆve Projects Jefferson Douglas At Risk Project
Completed Projects (FY) Kitsap Lincoln Spokane
King
Under Construcˆon No Current Projects
Mason
Grays Grant
Ki†tas
KPI Status Harbor Pierce Adams
TIB Commitment Level Thurston Whitman
Net Revenue Pacific Franklin
Lewis Garfield
Construcˆon Payments
Yakima Columbia
Average Payment Cycle Walla
Cowlitz Asoˆn
Delayed Projects Wahkiakum Skamania Benton Walla
Transacˆon Processing Klickitat
Clark

Rollover map for details

In this chapter, we discuss specific elements that can help create an effective data dash-
board. We begin by discussing what data dashboards are and what they can accomplish. We
then discuss various types of data dashboards, principles of good data dashboard design,
and characteristics of effective data dashboards. We show how to build a data dashboard
using Excel. We conclude the chapter by considering common errors made in data dash-
board design and strategies for avoiding these mistakes.

8-1 What Is a Data Dashboard?

In an automobile dashboard, values such as speed, fuel level, and oil pressure are displayed to
give the driver a quick overview of current operating characteristics of the vehicle. The driver can
use the information provided on the dashboard to operate the vehicle effectively and efficiently.
The engine coolant temperature gauge, fuel gauge, turn signal indicators, gearshift position indi-
cator, seat belt warning light, parking-brake warning light, emergency signals, and speedometer
provide such information. Other information on the dashboard, such as the odometer and oil
change indicator light, pertains to longer-term maintenance issues. Still other information, such
as the check engine light, aids in the identification of problems that need to be addressed more
immediately because they can cause serious or even irreparable damage.
 8-1 What Is a Data Dashboard? 325

Decision makers such as business managers have similar needs for information that will
enable them to operate and maintain their organizations effectively and efficiently. Such
Key performance indicators
are sometimes referred to
values may include the organization’s financial position, inventory on-hand, pending orders
as key performance metrics for raw materials, progress on projects, status of accounts, pending orders from customers,
(KPMs). and customer service metrics. These values are referred to as key performance indicators
(KPIs). In a healthcare setting, KPIs could refer to patient vital signs such as heart rate,
respiration rate, and blood pressure. For the manager of a political campaign, KPIs could
include fundraising values, recent polling results, and campaign expenditures.
Many decision makers rely on data dashboards to provide them with timely information on
their organization KPIs. A data dashboard is a data visualization tool that gives multiple out-
puts and may update in real time. The outputs provided by the dashboard are a set of KPIs for
the organization that are aligned with the organization’s goals and can be used to monitor current
and potential future performance on a continual basis. By consolidating and presenting data from
a number of sources in a data visualization designed for a specific set of purposes, data dash-
boards can help an organization better understand and use its data to improve decision making.

Principles of Effective Data Dashboards

Ideally, a data dashboard should present all KPIs related to some aspect of the organiza-
tion’s operations on a single screen that a user can easily review to understand the organi-
zation’s current state of operations quickly and accurately. Rather than requiring the user to
scroll vertically and horizontally to see the entire dashboard, it is better to create multiple
dashboards of related KPIs so that each dashboard can be viewed on a separate screen.
The KPIs displayed in the data dashboard should quickly and clearly convey mean-
ing to its user and be related to decisions the user makes. For example, the data dash-
board for a marketing manager may include KPIs related to current sales measures and
sales by region. The data dashboard for a foundation officer at a university may show
information on potential donors and their history of donations. The data dashboard for
a chief financial officer may provide information on the current financial standing of
the company. The data dashboard for a police chief may provide information on cur-
rent rates of crime in different neighborhoods.

Applications of Data Dashboards

Data dashboards are used in many different applications. In this section, we describe a few
common uses of dashboards in organizations.
Investment Dashboards — these dashboards are used to monitor the performance of
an organization’s or individual’s portfolio of investments overall and by individual invest-
ment and type of investment. An investment dashboard supports ongoing decisions on how
to best allocate invested funds, perhaps to maximize the rate of return while keeping risk at
or below a certain level or minimize risk while keeping the rate of return at or above a cer-
tain level. Information provided by these dashboards may include, but are not limited to:
●● current price, recent rate of return, and the relative rate of risk of the portfolio and its
individual components relative to the overall market
●● current price, recent rate of return, and relative risk of investments under consideration

●● share of the portfolio devoted to various industries or types of investments

●● key financial data for the companies in which the investor has invested or is consider-

ing investing
Manufacturing dashboards — these dashboards are used to monitor a production
process overall and by facility, product, individual, machine, shift, or department. A manu-
facturing dashboard supports ongoing decisions on how to allocate resources to produce an
organization’s products and/or services efficiently and effectively. Information provided by
these dashboards may include, among other metrics:
●● quantity produced
●● quality of units produced
326 Chapter 8 Data Dashboards

●● rate at which the units are produced

●● frequency of problems (machine downtime, returns, falling short of production
targets, etc.)
Marketing dashboards — these dashboards are used to monitor sales and promotional
efforts overall and by product or service, individual, campaign, division or department, or
customer. A marketing dashboard supports ongoing decisions on how to allocate resources
to design, price, promote, and distribute an organization’s products and/or services effi-
ciently and effectively. Information provided by these dashboards may include a variety of
metrics, for example:
●● quantity sold
●● quantity delivered
●● success rate/proportion of successful sales calls

●● sales revenue

●● prices at which products and services are sold

●● advertising and promotional expenditures

●● exposure and response to advertising and promotion efforts

●● contact with prospective customers

●● consumer satisfaction metrics

Within the category of marketing dashboards are more specific dashboards on areas such as
customer service, web page utilization, and social media effectiveness.
Human resource dashboards — these dashboards are used to monitor performance of
an organization’s workforce by individual, division, department, or shift. A human resource
dashboard supports ongoing decisions on how to allocate resources to ensure an organi-
zation utilizes its workforce in an efficient and effective manner. Information provided by
these dashboards may include, but are not limited to:
●● number of employees
●● length of employment
●● employee churn (rate of employee turnover)

●● reasons for leaving

●● employee performance

●● employee satisfaction

●● absenteeism

●● participation in training

●● overall labor effectiveness

Technical support dashboards — these dashboards are used to monitor performance of

an organization’s support of its technology users by individual, computer system, division or
department, or location. A technical support dashboard provides information to enhance ongo-
ing decisions on how to allocate resources to ensure an organization utilizes its workforce in an
efficient and effective manner. Information provided by these dashboards may include:
●● type of technical problem
●● response time
●● time to resolution

●● outstanding issues

●● persistent problems

●● downtime
Personal fitness/health
●● scheduled maintenance
dashboards are often
populated with data ●● software and hardware usage

collected by devices such

as smartwatches, fitness Personal fitness/health dashboards — these dashboards are used to monitor
monitors, and phones. aspects of an individual’s physical well-being. A personal fitness or health dashboard
8-2 Data Dashboards Taxonomies 327

supports ongoing decisions on how to achieve the best possible health for the individual.
Information provided by these dashboards may include, but is not limited to:
●● pulse
●● blood pressure
●● body temperature

●● weight and body-mass index (BMI)

●● amount of exercise

●● steps and distance travelled

●● calories burned

●● diet characteristics (calories, fat intake, carbohydrate intake, and protein intake)

●● blood sugar levels

Donor dashboards — these dashboards are used to monitor donor/contributor activity

and potential for nonprofit organizations by donor/contributor, prospective donor/contrib-
utor, or project/campaign. A donor dashboard supports ongoing decisions on how to use
the organization’s resources to maximize short- and long-term donor activity. Information
provided by these dashboards may include, for example:
●● number and dollar value of donations/contributions
●● contact with new potential donors/contributor and past donors/contributor
●● expenditures

●● conversion rate

●● cash flow

●● current status of campaigns

Crime dashboards — these dashboards are used to monitor the occurrences of differ-
ent types of crime in many cities. Crime dashboards can be used to inform the public about
the current and historical rates of crime and can be used by police and other government
administrators to make decisions on how best to deploy and utilize resources to increase
public safety. Information provided by these dashboards may include, for example:
●● number of police reports
●● number and dollar value of property crimes
●● number of arrests

●● types of crimes reported

School performance dashboards — these dashboards are used to monitor the perfor-
mance of schools. School performance dashboards can be used by parents to judge the rela-
tive performance of schools and by school administrators to aid in decisions related to staffing
and resource allocation. Information provided by these dashboards may include, for example:
●● number of students enrolled
●● demographic information on students
●● number of teachers and other staff employed

●● student test scores

●● number and types of courses offered

Data dashboards have been developed and successfully deployed for a wide variety of
other applications. Any organization that needs to quickly understand a related set of rap-
idly changing KPIs could benefit from a well-designed data dashboard.

8-2 Data Dashboards Taxonomies

Dashboards are generally classified by whether the information they provide is updated,
whether users can interact with the display, and what organizational functions they primar-
ily support. In this section, we describe these data dashboard taxonomies.
328 Chapter 8 Data Dashboards

Data Updates
A data dashboard can be classified into one of two groups based on how often the informa-
tion it provides is updated. A static dashboard provides information on an organization’s
KPIs that may periodically be updated manually as new data and information are collected.
These dashboards are relatively inexpensive and easy to develop, are generally updated
infrequently, and are useful when the organization’s KPIs change slowly.
A dynamic dashboard provides information on an organization’s KPIs and regu-
larly receives and incorporates new and revised data and incorporates these data into the
dashboard. These dashboards take more time and effort to develop, are generally updated
frequently (perhaps continuously), and are useful when the organization’s KPIs change
rapidly.

User Interaction
A data dashboard can be classified into one of two groups based on whether users can cus-
tomize ther displays. A noninteractive dashboard does not allow users to customize the
data dashboard display. These dashboards are useful when the data on which the dashboard
is based do not change frequently.
Although interactive Conversely, an interactive dashboard allows users to customize the data dashboard
dashboards can be either display, effectively allowing a user to filter the data displayed to the user on the dashboard.
static or dynamic, they are
Dashboards can allow for user interactivity in a variety of ways, including:
generally dynamic.
●● Drilling down — a feature that provides the user with more specific and detailed
information on a particular element, variable, or KPI. A drill-down can take the user
to a new display with additional detailed information when the user clicks on a par-
ticular element, variable, or KPI. It can also provide a popup display with additional
detailed information when the user rolls the cursor over a particular element, vari-
able, or KPI.
●● Hierarchical filtering — a feature that provides the user with the capability to

restrict the data displayed to a specific segment of users by systematically selecting

values of several categories or values of variables in a nested manner, for example,
filter the data displayed by first selecting a gender, then selecting an age group, then
selecting the product purchased, etc.
●● Time interval widget — a feature that allows the user to specify the time period to

be displayed on the dashboard.

●● Customization tools — features that allow the user to

❯❯ select/deselect categories to be displayed on the dashboard

❯❯ combine categories to be displayed on the dashboard

❯❯ select/deselect variables to be displayed on the dashboard

❯❯ add custom text and labels to the dashboard

❯❯ add or hide various summary statistics for variables on the dashboard

❯❯ hyperlink a particular element, variable, or KPI to related content (even an external

website or resource)
Interactive dashboards can provide the user a great deal of latitude in exploring the data
provided by the dashboard, finding specific relevant information, and independently learn-
ing about and finding potential solutions to the problem she or he is attempting to resolve.

Organizational Function
A data dashboard can be classified into one of four groups based on its purposes and its
ultimate users.
Operational dashboards are typically used by lower level managers to monitor rapidly
changing critical business conditions. Because these data usually accumulate swiftly and
are critical for the daily operations of the organization, these dashboards generally update
in real time or multiple times throughout the day.
8-3 Data Dashboard Design 329

Tactical dashboards are typically used by mid-level managers to identify and assess the
organization’s strengths and weaknesses in support of the development of organizational strat-
egies. Because tactical dashboards usually support the development of organizational strate-
gies, these dashboards are generally updated less frequently than an operational dashboard.
Strategic dashboards are typically used by executives to monitor the status of KPIs’ rele-
vant overarching organizational objectives. The data that support a strategic dashboard update
on a recurring basis but at less frequent intervals than tactical and operational dashboards.
Analytical dashboards are typically used by analysts to identify and investigate trends,
predict outcomes, and discover insights in large volumes of data.
Some businesses use dashboards that span more than one of these categories with great
success, and many organizations have developed and use dashboards that do not fall into
any of these four categories. When designing a data dashboard, it is important to consider
how these various types of data dashboards support the objectives of the organization for
which you are designing the dashboard and meet the needs of the dashboard’s end users.
We explore this issue further in the next section.

8-3 Data Dashboard Design

Although there is a great deal of variation in the design of data dashboards, several consid-
erations are common to successful dashboard design. In this section, we provide an over-
view of these considerations.

Understanding the Purpose of the Data Dashboard

At this stage, the organization assesses what the proposed data dashboard should do for its
users and whether a data dashboard is the appropriate tool. The primary reason for develop-
ing a data dashboard is to support an organization’s operations, decision making, and stra-
tegic planning, so the overriding consideration in designing a data dashboard is its ultimate
purpose: What does the organization need to accomplish through use of the dashboard once it
is operational? An organization’s objectives for its data dashboard may include, for example:
●● tracking KPIs
●● monitoring processes
●● assessing attainment of goals and objectives

●● developing/enhancing insight

●● sharing information

●● measuring performance

●● forecasting

●● data exploration

Failure to consider the organization’s motivations for creating a dashboard will leave the dash-
board design team directionless, which can slow the development of the dashboard and poten-
tially result in development of a dashboard that does not address the needs of the organization.

Considering the Needs of the Data Dashboard’s Users

A data dashboard should be designed to assist a particular user or group of users with specific
tasks associated with the management of an organization. Thus, it is important that the dash-
board design team both understands the needs of the dashboard’s end users and recognizes
how addressing these needs will ultimately support the organization’s dashboard objectives.
This knowledge, in conjunction with adherence to the principles of effective data visualiza-
tion, will help the dashboard developer determine the information that the dashboard should
convey and the most effective manner for presenting this information to its intended audience.

Data Dashboard Engineering

Once the dashboard design team understands the organization’s objectives for the dashboard
and the dashboard end users’ related needs, the team should turn its attention to the information
330 Chapter 8 Data Dashboards

that should displayed in the dashboard. All information displayed must meet the end users’
needs, and the design dashboard team should work with the end users to ensure this occurs.
Selection of an appropriate After the information to be displayed is identified and organized, the dashboard design
type of chart, effective use team should determine the manner in which the information will be displayed. This includes
of preattentive attributes
selection of appropriate types of charts, effective use of preattentive attributes and Gestalt
and Gestalt principles, and
appropriate use of color are
principles, appropriate use of color, and an effective layout that enables end users to easily
discussed in Chapters 2, 3, find the information they need and relate information from various charts in the dashboard.
and 4, respectively It is also important that the dashboard is easy to read and interpret, and that the display
is not too sparse, too crowded, or overly complex. Strategies for avoiding overcrowding
and unnecessary complexity include:
●● avoiding inclusion of information that will not be useful to the end users
●● organizing the information into subsets that each address a different need of the end
users and displaying information in these subsets across multiple pages
●● using interactive dashboard tools (drilling down, hierarchical filtering, time interval

widget, customization tools)

In addition, at this stage the dashboard design team should consider the environment in
which the data dashboard will be used. Data dashboards are most often accessed from
desktop computers. However, some data dashboards are accessed with other devices such
as tablets or smartphones on factory floors or retail showrooms, in automobiles, or out-
doors. It is important to consider factors such as the device used to access the dashboard,
ambient lighting, size and resolution of the display, likely distance of the user from the
display, and whether a touch screen will be utilized when designing a data dashboard.
The design team should consider the most effective ways to provide context for the infor-
mation presented in the data dashboard. This can be done in many ways, including:
●● showing how a KPI varies over time
●● comparing the value of a KPI to an organizational goal
●● comparing the value of a KPI

❯❯ internally across divisions, departments, or geographies of an organization

❯❯ externally across customers or market segments

❯❯ externally across competitors or organizations in the same industry

By providing appropriate context, the dashboard gives meaning to the data and enhances
the dashboard user’s ability to interpret and act on the data. Throughout this step, it is
important that the dashboard design team collaborate with the dashboard’s end users
to ensure the dashboard meets their needs and doesn’t incorporate extemporaneous
information.
The data-ink ratio and The dashboard design team must understand how the data dashboard will be used so it can
reducing eye travel are organize the charts and tables on the dashboard in a way that facilitates analyses by the users.
discussed in Chapter 3.
When designing the individual components of the data dashboard, the design team should be
mindful of the data-ink ratio and reducing eye travel. The dashboard design team must also
understand how the data dashboard will be maintained and how the data dashboard’s effec-
tiveness will be assessed. A well-designed data dashboard can quickly lose its value to its
users if it is difficult to update and maintain, so the design team should consider the skills and
capabilities of the individual or team that will be responsible for its maintenance and updates.
Although the current sources of data and their format are critical considerations in dash-
board design, the dashboard design team should also reflect on the future of the dashboard
and discuss this issue with management. Could the organization’s objectives shift in the
future? If so, how will the data dashboard need to reflect these shifts? What new KPIs are
likely to become important to the organization in the future? What may be the source or
format of new data to be incorporated into the dashboard in the future?
Finally, errors, miscommunications, and misunderstandings occur in virtually all
complex projects. These problems can result in the creation of a faulty final product that
can lead to poor decisions and missed opportunities, require time-consuming and costly
 8-4  Using Excel Tools to Build a Data Dashboard 331

revisions, and damage the credibility of the data dashboard design team. It is crucial that
the dashboard design team also test its work extensively at each step to ensure the dash-
board functions as the team intends. At critical junctures, the dashboard design team must
review its progress with the management team that initiated the development of the data
dashboard to ensure the dashboard under production is meeting the management team’s
objectives. Finally, the data dashboard design team must provide the ultimate users with
opportunities to test the dashboard extensively to ensure the dashboard is user friendly,
functions in the manner expected by the users, and produces the outputs needed by its
users. By following a process that adheres to these guidelines, malfunctions can be identi-
fied during the development of the data dashboard when they are easier, less expensive, and
less time consuming to correct.

8-4 Using Excel Tools to Build a Data Dashboard

We have already reviewed several tools that are useful for constructing data dashboards.
PivotTables, PivotCharts, A PivotTable is an interactive crosstabulation that allows the user to interact with the
and slicers are discussed in data by applying filters to select various aspects of the data to be displayed in the table.
Chapter 6.
A PivotChart is a chart that allows the user to interact with the data by applying filters
to select various aspects of the data to be displayed in the chart. A slicer allows the user
to filter the data to be displayed in PivotTables and PivotCharts. Because these tools
provide users with the capability of interacting with tables and charts and selecting
what is to be displayed, well-designed PivotTables, PivotCharts, and slicers allow the
user to delve deeper into the data and learn from the data in a more focused manner.
In general, the original source of data for the dashboard should be stored separately from
the dashboard, and the dashboard should not draw information directly from the original
source of data. In other words, the data that are needed to create an individual chart or table
on the data dashboard should be extracted from the original source data, and they should be
stored separately. The dashboard’s users should not be able to permanently change the dash-
board. In a spreadsheet, this means that the original source of data for the dashboard should
be stored in its own worksheet, and the dashboard should not draw information directly
from that worksheet. The data for each unique display unit (table, chart, etc.) in the dash-
board should be maintained in a separate worksheet. And the components of the dashboard
should be locked so the user cannot make permanent changes.

Espléndido Jugo y Batido, Inc.

We introduced the reader to Let us again consider Espléndido Jugo y Batido, Inc. (EJB), a company that bottles
Espléndido Jugo y Batido, Inc juices and smoothies in five fruit flavors (apple, grape, orange, pear, and tomato) and
in Chapter 6.
four vegetable flavors (beet, carrot, celery, and cucumber). EJB now wants to develop a
data dashboard that its mid-level operations management team can use to track its sales
over the most recent three years. By working through the data dashboard design process
discussed in the previous section, the company’s management team has established its
objectives for this dashboard. Specifically, EJB wants to be able to track dollar sales for
each of its distribution centers by year, and it wants to be able to look at these data for
new and existing customers. EJB also wants to be able to track its dollar sales by year
for each category (juices and smoothies), and it would like to be able to look at these
data by flavor and by month. The company wants to monitor time to deliver by distribu-
tion center for each year, and it wants to be able to generate sales in dollars across cate-
gory and flavor by distribution center at the year, month and date ordered levels. Finally,
EJB wants this data dashboard to automatically update when new data are added to the
source data.
Thus, EJB’s KPIs for this dashboard are total dollar sales and average delivery times. In
addition, EJB’s desire to be able to specify the month ordered, year ordered, date ordered,
category, flavor, distribution center, and whether new or existing customers are displayed in
various charts indicates that the company needs a dynamic and interactive data dashboard.
332 Chapter 8 Data Dashboards

Now that the purpose of the data dashboard, the relevant KPIs, the objectives for creat-
ing the data dashboard, and the needs of the dashboard’s users have been considered, we
are able to determine that we can provide EJB the functionality it desires in its data dash-
board through the following charts and tables:
●● a stacked column chart of total sales across distribution centers and year ordered by
new or existing customers
●● a line chart of total sales across year and month ordered by new or existing customers
EJBChart1
●● a clustered column chart of total sales across year ordered and flavor by category

●● a clustered bar chart of average time to deliver across distribution centers by year

ordered
●● a table of total sales across category and flavor by distribution center and year,

month, and day ordered

If you click the PivotChart to

open the PivotTable Fields Using PivotTables, PivotCharts, and Slicers to Build
pane, the Rows area will a Data Dashboard
be referred to as the Axis
(Categories) area and the We will use the file EJBChart1, which contains the EJBData Excel table in the Data
Columns area will be referred worksheet with the fields we need and the chart and slicers provided in Figure 8.3 in the
to as the Legend (Series) area. Chart1 worksheet to create the remaining components of the EJB data dashboard.

FIGURE 8.3 PivotChart for Total Sales with Slicers for Distribution Center, Year Ordered,
and New Customer
 8-4  Using Excel Tools to Build a Data Dashboard 333

Recall that you can select Note that this PivotChart will allow the user to display total dollar sales for any combi-
multiple items in a slicer by
nation of years ordered, distribution centers, and whether or not the customers are new. For
clicking a slicer button while
holding the Ctrl key, then
example, suppose the user wants to compare sales to existing customers for the Idaho (ID) and
clicking on each additional Rhode Island (RI) distribution centers in 2019 and 2020. The user can create a chart for this
item you wish to select in purpose by selecting ID and RI in the Distribution Center slicer, 2019 and 2020 in the Year
that slicer. Ordered slicer, and No in the New Customer? slicer. This produces the chart in Figure 8.4.

FIGURE 8.4 PivotChart for Total Sales to Existing Customers by the Idaho and Rhode Island
Distribution Centers in 2019 and 2020

This chart shows that total sales to existing customers increased substantially from 2019
to 2020 at both the Idaho and Rhode Island distribution centers.
Also note that Excel’s tooltip feature allows the user to open a popup window with addi-
tional information for a portion of a table or chart by hovering the cursor over a portion of
an Excel table or chart as shown in Figure 8.5. These popups will be active on the charts
that we include in a dashboard. They provide users with some drill-down capability and
can be customized to deliver a wide variety of information.
We can use PivotTables, PivotCharts, slicers, and the steps outlined in Chapter 6 to build
the remaining three charts
●● a line chart of total sales across year and month ordered by a new or existing customer
●● a clustered column chart of sales across year ordered and flavor by category
●● a clustered bar chart of average time to delivery across distribution centers by year

ordered for EJB’s data dashboard.

334 Chapter 8 Data Dashboards

FIGURE 8.5 Popup Window with Additional Information on 2020 Idaho New Customers

Excel Tables are discussed in By creating an Excel table in the EJBData worksheet and using this Excel table as the
Chapter 6. source data for PivotTables and PivotCharts, we greatly simplify the process of updating
the data, PivotCharts, and PivotTables. To add a new record to the existing data, we now
only need to enter the record into the row adjacent to the last row of the EJBData table.
To add a new field to the existing data, we only have to enter the field into the column adja-
cent to the last column of the EJBData table; Excel will automatically incorporate this new
information into the EJBData table. To delete a record a field from the existing data, we
now only have to delete the entire corresponding row or column from the table.
Once new data have been added to the EJBData table, we can quickly refresh associated
PivotTables and PivotCharts to reflect the new data (or any revisions to existing data) by
clicking anywhere in a PivotTable, clicking the PivotTable Analyze tab on the Ribbon,
clicking the Refresh button in the Data group, and clicking Refresh to update the
selected PivotTable or Refresh All to update all PivotTables in the file simultaneously.
Thus, we have created a dynamic data dashboard that can be refreshed quickly to reflect
recently added and revised data without updating the data range originally selected.
To create the line chart of total sales across year and month ordered by category, we use
the following steps.
Step 1. In the file EJBChart1, select any cell in the EJBData table in the Data worksheet
Click the Insert tab on the Ribbon
In the Charts group click PivotChart and then select PivotChart &
PivotTable
 8-4  Using Excel Tools to Build a Data Dashboard 335

Step 2. When the Create PivotTable dialog box appears:

Under Choose the data that you want to analyze, choose Select a table
or range and enter EJBData in the Table/Range: box
Under Choose where you want the PivotTable report to be placed,
select New Worksheet
Click OK
Step 3. Change the name of the new worksheet to Chart2
We next create the line chart of total sales by month and year by category in the Chart2
worksheet.
If the PivotTable Fields task Step 4. In the PivotChart Fields task pane:
pane does not appear, right Drag the $ Sales field to the Values area
click in the PivotTable and
Drag the Year Ordered and Month Ordered fields to the Axis (Categories)
select Show Field List.
area
Drag the Category field to the Legend (Series) area
Use the Value Field Settings from the drop down menu of $ Sales to
select Sum for the $ Sales field
Step 5. Click the chart, click the Design tab on the PivotChart Tools Ribbon, click the

Change Chart Type button Change

in the Type group, click Line Line on
Chart Type

the left pane, and select Line in the right pane

Click OK
Step 6. Right click the vertical axis of the chart and click Format Axis… to open the
Format Axis task pane
Click Axis Options and then click the Axis Options button
Under Axis Options, enter 0 in the Minimum box
In the Display Units box, select Thousands
Click on the vertical axis title of “Thousands” that has automatically been
generated on the chart and press the Delete key
Step 7. Right click the horizontal axis and click Format Axis… to open the Format
Axis task pane
Click Axis Options and then click the Axis Options button
Click Labels, select Specify interval unit, and enter 1
Step 8. Right click any Field Button (such as Month Ordered ) and select Hide All
Field Buttons on Chart
Step 9. Click the PivotChart, click the Insert tab on the Ribbon, and click Slicer
in the Filters group Slicer

Step 10. When the Insert Slicers dialog box appears:

Select the check boxes for Category, Year Ordered, and Month
Ordered
Click OK
For the chart shown in
Step 11. Click each of the slicers and use the tools in the Slicer tab on the Ribbon to
Figure 8.6 the borders have
been removed from the slicers format the slicer as appropriate
and the number of columns Drag the slicers to the positions they should occupy and the dashboard
has been set to 2 for the and resize the slicers accordingly
Category slicer, 3 for the Year
Ordered slicer, and 6 for the Some additional editing for readability to the chart created using the preceding steps results
Month Ordered slicer. in the chart shown in Figure 8.6, which can also be found in the Chart2 worksheet in the
file EJBCharts. This PivotChart will allow the user to display total sales for any combina-
tion of years ordered, months ordered, and categories.
To create the clustered column chart of total sales across year ordered and flavor by cat-
EJBCharts
egory, we follow these steps.
336 Chapter 8 Data Dashboards

FIGURE 8.6 PivotChart for Total Sales with Slicers for Year Ordered, Month Ordered, and New
Customer

Step 1. In the file EJBChart1, select any cell in the EJBData table in the Data
worksheet
Click the Insert tab on the Ribbon
In the Charts group click PivotChart and then select PivotChart &
PivotTable
Step 2. When the Create PivotTable dialog box appears:
Under Choose the data that you want to analyze, select Select a table
or range and enter EJBData in the Table/Range: box
Under Choose where you want the PivotTable report to be placed,
select New Worksheet
Click OK
Step 3. Change the name of the new worksheet to Chart3
Step 4. In the PivotChart Fields task pane:
Drag the $ Sales field to the Values area
Drag the Year Ordered and Flavor fields to the Axis (Categories) area
Drag the Category field to the Legend (Series) area
 8-4  Using Excel Tools to Build a Data Dashboard 337

Use the Value Field Settings from the drop down menu of $ Sales to
select Sum for the $ Sales field
Step 5. Click the chart, click the Design tab on the PivotChart Tools Ribbon, click the

Change Chart Type button Change

in the Type group, click Column on the left
Chart Type

pane, and select Clustered Chart in the right pane

Click OK
Step 6. Right click the vertical axis of the chart and click Format Axis… to open the
Format Axis task pane
Click Axis Options and then click the Axis Options button
Under Axis Options, enter 0 in the Minimum box
In the Display Units box, select Thousands
Click on the vertical axis title of “Thousands” that has automatically been
generated on the chart and press the Delete key
Step 7. Right click any Field Button (such as Month Ordered ) and select Hide All
Field Buttons on Chart
Step 8. Click the PivotChart, click the Insert tab on the Ribbon, and click Slicer
Slicer
in the Filters group
Step 9. When the Insert Slicers dialog box appears:
Select the check boxes for Category, Year Ordered, and Month
Ordered
Click OK
For the chart shown in Step 10. Click each of the slicers and use the tools in the Slicer tab on the Ribbon to
Figure 8.6 the borders have format the slicer as appropriate
been removed from the slicers
Drag the slicers to the positions they should occupy and the dashboard
and the number of columns
has been set to 2 for the
and resize the slicers accordingly
Category slicer, 3 for the Year
Some additional editing for readability to the chart created using the preceding steps results
Ordered slicer, and 4 for the
Flavor slicer.
in the chart shown in Figure 8.7, which can also be found in the Chart3 worksheet in the
file EJBCharts.
This PivotChart will allow the user to quickly display total sales for any combination of
categories, flavors, and years ordered.
To create a clustered bar chart of total sales across distribution center by year ordered,
EJBCharts
we follow these steps.
Step 1. In the file EJBChart1, select any cell in the EJBData table in the Data worksheet
Click the Insert tab on the Ribbon
In the Charts group click PivotChart and then select PivotChart &
PivotTable
Step 2. When the Create PivotTable dialog box appears:
Under Choose the data that you want to analyze, choose Select a table
or range and enter EJBData in the Table/Range: box
Under Choose where you want the PivotTable report to be placed,
select New Worksheet
Click OK
Step 3. Change the name of the new worksheet to Chart4
We next create the clustered bar chart of total sales across distribution center by year
ordered in the Chart4 worksheet.
Step 4. In the PivotChart Fields task pane:
Drag the Time to Deliver field to the Values area
Drag the Distribution Center field to the Axis (Categories) area
Drag the Year Ordered field to the Legend (Series) area
Use the Value Field Settings from the drop down menu of Time to
Deliver to select Average for the Time to Deliver field
338 Chapter 8 Data Dashboards

FIGURE 8.7 PivotChart for Total Sales with Slicers for Year Ordered, Flavor, and Category

Step 5. Click the chart, click the Design tab on the PivotChart Tools Ribbon, click the

Change Chart Type button Change

in the Type group, click Bar Bar on
Chart Type

the left pane, and select Clustered Bar in the right pane
Click OK
Step 6. Right click the vertical axis and click Format Axis to open the Format Axis pane
Click Axis Options and then click the Axis Options button
Click Labels, click Specify interval unit, and enter 1 into
the Specify interval unit box
Click OK
Step 7. Right click any Field Button (such as Year Ordered ) and select Hide All
Field Buttons on Chart
Step 8. Click the PivotChart, click the Insert tab on the Ribbon, and click Slicer
Slicer
in the Filters group
 8-4  Using Excel Tools to Build a Data Dashboard 339

For the chart shown in Figure Step 9. When the Insert Slicers dialog box appears:
8.6 the borders have been
Select the check boxes for Distribution Center and Year Ordered
removed from the slicers
and the number of columns
Click OK
has been set to 7 for the Step 10. Click each of the slicers and use the tools in the Slicer tab on the Ribbon to
Distribution Center slicer, and format the slicer as appropriate
3 for the Year Ordered slicer. Drag the slicers to the positions they should occupy and the dashboard
and resize the slicers accordingly
Some additional editing for readability to the chart created using the preceding steps results
in the chart shown in Figure 8.8, which can also be found in the Chart4 worksheet in the
file EJBCharts. This PivotChart will allow the user to quickly display average time to
EJBCharts
deliver for any combination of distribution centers and years ordered.

FIGURE 8.8 PivotChart for Average Time to Deliver with Slicers for Distribution Center and Year
Ordered

We now turn our attention to developing the PivotTable to be included in the EJB data
dashboard. To create the PivotTable of total dollar sales and average time to deliver across
category and flavor by distribution center and year, month, and day ordered, we follow
these steps.
340 Chapter 8 Data Dashboards

Step 1. In the file EJBCharts, create a new worksheet and name this worksheet
Dashboard
Step 2 changes the fill color of cells A1:Z100 to a dark color to provide a contrasting back-
ground for the dashboard.
Step 2. Select cells A1:Z100
Click the Home tab on the Ribbon and in the Font group, click the Fill
Color button and select a dark blue color
Step 3. In the file EJBCharts, select any cell in the EJBData table in the Data
worksheet
Click the Insert tab on the Ribbon and select PivotTable in the Tables
group
Step 4. When the Create PivotTable dialog box appears:
Under Choose the data that you want to analyze, click Select a Table
or Range and enter EJBData in the Table/Range: box
Under Choose where you want the PivotTable report to be placed,
select Existing Worksheet and enter Dashboard!B4
Click OK
Step 5. In the Dashboard worksheet
Click the empty PivotTable to open the PivotTable Fields task pane
Step 6. In the PivotTable Fields pane:
Drag the $ Sales and Time to Deliver fields to the Values area
Drag the Year Ordered, Month Ordered, Day Ordered, and
Distribution Center fields to the Filters area
Drag the Category field to the Columns area (make sure this is the first
field listed in the Columns area)
Drag the Flavor field to the Rows area
Click the drop down arrow next to Time to Deliver in the Values area,
click Value Field Settings... and change the Summarize value field by
to Average
Click OK
Once you have completed this step, the Drag fields between areas below: area in the
PivotTable Fields pane should look like Figure 8.9. Note that the order in which the fields
are listed in each area determines the layout of the resulting PivotTable.

FIGURE 8.9 The Drag fields between areas below: Area for the EJB
PivotTable
 8-4  Using Excel Tools to Build a Data Dashboard 341

Step 7. Right click on any cell in the PivotTable and select PivotTable Options...
Step 8. When the PivotTable Options dialog box opens:
Click the Layout & Format tab
Deselect the check box for Autofit column widths on update
This creates the PivotTable shown in Figure 8.10, which can also be found in the
Dashboard worksheet in the file EJBDashboard1.
EJBDashboard1

FIGURE 8.10 PivotTable for EJB Data Dashboard

This PivotTable will allow the user to find total dollar sales and average time to deliver
for any combination of categories, flavors, and distribution centers for any years, months,
or dates ordered.
Now that we have created each of the data dashboard’s components, we can assemble
the data dashboard. We start by considering the relative positioning of each of these charts
on the dashboard. our objective is to minimize eye travel by putting charts that are likely to
be used together in proximity to each other.
●● The chart of total sales by distribution center, year ordered, and new customer? (in
the Chart1 worksheet) and the chart of total sales (thousand $) by year ordered, fla-
vor, and category (in the Chart3 worksheet) will both be used in analyses of EJB’s
sales history by year and should be adjacent to each other.
●● The chart of total sales by distribution center, year ordered, and new customer? (in

the Chart1 worksheet) and the chart of average time to deliver by distribution center
and year ordered (in the Chart4 worksheet) will both be used in analyses at the distri-
bution center level and should be adjacent to each other.
●● The chart of total sales by year ordered, month ordered, and category (in the Chart2

worksheet) and the chart of total sales by year ordered, flavor, and category (in the
Chart3 worksheet) will both be used in analyses at the category level and should be
adjacent to each other.
●● The table of total sales across category and flavor by distribution center and

year, month, and day ordered should be at the top of the dashboard for easy
access.
342 Chapter 8 Data Dashboards

One configuration that satisfies these criteria is provided in Figure 8.11.

FIGURE 8.11 Potential Layout of EJB Data Dashboard

Table - Total Sales and Average Time to Deliver

Across Category and Flavor by Distribution Center and Year, Month,
and Day Ordered

Chart 3 - Total Sales Chart 2 - Total Sales

by Year by Year
Ordered, Flavor, and Ordered, Month Ordered,
Category and Category

Chart 1 - Total Sales

Chart 4 - Average Time to
by Distribution
Deliver by Distribution
Center, Year Ordered, and
Center and Year Ordered
New Customer

We will complete construction of the EJB data dashboard in this layout using the fol-
lowing steps.
Step 1. Click the PivotChart in Chart1 worksheet
Step 2. Click the Analyze tab on the PivotChart Tools Ribbon, and click the Move Chart
button Move
in the Actions group
Chart

Step 3. When the Move Chart dialog box appears:

Select Object in: and choose Dashboard
Click OK
Step 4: In the Chart1 worksheet, click on the Distribution Center slicer
Click the Home tab on the Ribbon
Click Cut in the Clipboard group
Step 5. In the Dashboard worksheet
Click the Home tab on the Ribbon
Click Paste
Step 6. Position the PivotChart and slicer where you want them on the dashboard
Step 7. Repeat Steps 1–6 to move the charts and slicers in the Chart2, Chart3, and
Chart4 worksheets to the Dashboard worksheet and position these objects
where you want them on the dashboard
Step 8. Double click one of the PivotCharts in the Dashboard worksheet to open the
Format Chart Area task pane
Click Chart Options, then click the Size & Properties button
Click Properties, then select Don’t move or size with cells
Step 9. Repeat Step 8 for the remaining charts on the Dashboard worksheet
Step 10. Right click one of the slicers in the Dashboard worksheet, then select Size
and Properties... to open the Format Slicer dialog box
Click Properties, then select Don’t move or size with cells
 8-4  Using Excel Tools to Build a Data Dashboard 343

Step 11. Repeat Step 10 for the remaining slicers on the Dashboard worksheet
Step 12. Right click any chart on the Dashboard worksheet and select PivotChart
Options...
Step 13. When the PivotTable Options dialog box opens:
Click the Layout & Format tab
Deselect the check box for Autofit column widths on update
Click OK
Step 14. Adjust the positioning and size of the charts and slicers so they align as desired
Step 15. Select the View tab on the Ribbon, and in the Show group deselect the check
box for Gridlines
The dashboard created to this point looks great, but when we use a slicer to filter out some
series on a chart, the colors assigned to the remaining series may change. Furthermore, this
dashboard uses blue and orange to differentiate between new and existing customers, juices
and smoothies, and 2018 and 2019, if we use different sets of colors to consistently differen-
tiate between each of these series across charts. The following steps address both concerns.
Step 16. Click the File tab on the Ribbon, then select Options
Step 17. When the Excel Options dialog box appears:
Click Advanced
Use Format Data Series to
Under Chart, select the check boxes for Properties follow chart data
check each series on each point for all new workbooks and Properties follow chart data point
chart to ensure that none of for current workbook
the colors assigned to the Step 18. Assign the specific colors below to different chart elements by right clicking
series in any of the charts in
on that chart element, clicking Fill and selecting More Fill Colors...
the dashboard are reassigned
automatically by Excel when
Blue to existing customers and Orange to new customers
the slicers are used to filter Purple to juices and red to smoothies
the data. Brown to 2018, Pink to 2019, and Green to 2020
Step 19. Change the fill color of cells A1:AL80 to dark blue to create a contrasting
background
Step 20. Type Espléndido Jugo y Batido, Inc. Sales Dashboard into cell A2
Change the font to white 48 pt. Brush Script MT font
Step 21. Select cells A2:AF2, select the Home tab on the Ribbon, and in the
Alignment group click Merge & Center
Some additional editing for readability to the table created using the preceding steps results
EJBDashboard1 in the dashboard shown in Figure 8.12, which can also be found in the Dashboard work-
sheet in the file EJBDashboard1.
You can change the style and settings for any slicers by clicking the slicer, clicking
Options on the ribbon, the using the various tools such as those in the Slicer Styles group.

Linking Slicers to Multiple PivotTables

The data dashboard in Figure 8.12 allows the user to filter each chart separately using the
associated slicer(s). This is a definite advantage of this dashboard design for users who
need this level of control over the charts on their dashboard. However, users may prefer to
use a single slicer that controls multiple related charts. For example, a single Distribution
Center slicer that simultaneously controls the Distribution Centers that are displayed on all
relevant charts on the dashboard might be preferable.
We will illustrate with Distribution Center. Start by deleting all but one of the Distribution
Center slicers from the EJB dashboard as shown in Figure 8.13. The following steps will connect
the remaining Distribution Center slicer to the PivotTables in the Chart1 and Chart4 worksheets.
Step 1. Select the remaining Distribution Center slicer on the Dashboard worksheet
Click the Slicer tab on the Ribbon
Click the Report Connections button in the Slicer group
Report
Connections
344 Chapter 8 Data Dashboards

FIGURE 8.12 EJB Data Dashboard

 8-4  Using Excel Tools to Build a Data Dashboard 345

FIGURE 8.13 EJB Data Dashboard after Deleting All but One of the Distribution Center Slicers
346 Chapter 8 Data Dashboards

Step 2. Select both the PivotTables created in the Chart1 worksheet and the Chart4
worksheet
Click OK
Because the PivotCharts created in the Chart1 and Chart4 worksheets are linked to the corre-
sponding PivotTables in the Chart1 and Chart4 worksheets, the single Distribution Center
slicer now controls the value(s) of Distribution Center that are displayed in both of these charts.
Step 3. Repeat Steps 1 and 2 to link the PivotTables created in the Chart2 and Chart3
worksheets to a single Category slicer and to link the PivotTables created in the
Chart1, Chart2, Chart3, and Chart4 worksheets to a single Year Ordered slicer
With fewer slicers to display, we can rearrange and resize the charts in a manner that is
more visually appealing and easier to read while still meeting the criteria we established for
which charts should be adjacent to each other, leading to the result shown in Figure 8.14.

FIGURE 8.14 EJB Data Dashboard with Connected Slicers

In addition to allowing for simultaneous filtering of related charts and providing EJB
with easy access to the information it desires, the dashboard now appears less cluttered.
Note, however, that only PivotTables that are generated from the same data can be fil-
tered simultaneously by a single slicer. For the EJB data dashboard, we generated each
PivotTable from the EJBData table on the Data worksheet to ensure that any slicer we cre-
ate could be used to filter every chart we created.

Protecting a Data Dashboard

Finally, we need to prevent users from altering the dashboard while still allowing them to
interact with the dashboard through the slicers and filters. We will do so through the fol-
lowing steps.
Step 1. Right click any slicer on the Dashboard worksheet, then click Size and
Properties...
Step 2. When the Format Slicer task pane appears:
Under Position and Layout select the check box for Disable resizing
After selecting one slicer, you
and moving
can hold down the Ctrl key
and select additional slicers
Under Properties select Don't move or size with cells and deselect the
to execute steps 1 and 2 on check box for Locked
multiple slicers simultaneously Step 3. Repeat Steps 1 and 2 for each remaining slicer on the Dashboard worksheet
 8-4 Using Excel Tools to Build a Data Dashboard 347

Step 4. Select the range of cells that contains the PivotTable in the Dashboard work-
sheet (this is cells B4:H21 on our Dashboard worksheet)
Click the Review tab on the Ribbon
In the Protect group, click the Allow Edit Ranges button
Step 5. When the Allow Users to Edit Ranges dialog box appears:
Click the New... button
Step 6. When the New Range dialog box appears:
Enter DashboardPivotTable in the Title: box
Click OK
Step 7. When the Allow Users to Edit Ranges dialog box returns:
Click the Protect Sheet... button
The Protect Sheet button Step 8. When the Protect Sheet dialog box opens:
toggles to the Unprotect Deselect the check box for Select locked cells
Sheet button when the sheet
Select the check box for Use PivotTable & PivotChart
is protected and back to the
Protect Sheet button when
Enter a password in the Password to unprotect sheet: box
the sheet is not protected. Renter the same password in the Reenter password to proceed. box
of the Confirm Password dialog box (recall that we used the password
TRIAL to protect the dashboard in the file EJBDashboard2)
Click OK
Step 9. Select the Chart1 worksheet tab, then hold the Ctrl key down and select the
Chart2, Chart3, Chart4, and Data worksheet tabs so all worksheet tabs in the
EJB2Dashboard file except the Dashboard worksheet are selected
Right click the Chart1 worksheet tab and select Hide
The complete dashboard with linked slicers is in the Dashboard worksheet of the file
EJBDashboard2.
A user can now use the slicers and filters on the dashboard but cannot otherwise alter or
EJBDashboard2
move the PivotCharts, the PivotTable, or the slicers. Each of the worksheets other than the
Dashboard worksheet is also hidden from the user, preventing unwanted changes from being
made to the raw data and PivotCharts that have been used to create the data dashboard.
When you need to revise the dashboard, use the following steps to unprotect the
Dashboard worksheet and unhide the Chart1, Chart2, Chart3, Chart4, and Data worksheets.
Step 1. Right-click the Dashboard worksheet tab
Step 2. Click Unhide...
Select the hidden worksheet you wish to unhide

Final Review of a Data Dashboard

The initial data dashboard likely does not perform exactly as the design team intends, the
management team expects, or the users need. Rounds of reviews must again be completed
to ensure the final product works properly from the perspectives of all the dashboard’s
stakeholders:
●● The data dashboard design team must test its final work extensively. The slicers and
filters should be used at many levels and the results in the PivotCharts and PivotTa-
bles should be reviewed at these levels to ensure the information is accurate and the
display is appealing and informative.
●● The design team must also review the final data dashboard with the management

team that authorized the development of the data dashboard to ensure the data dash-
board meets the management team’s objectives.
●● The design team must provide the ultimate users with opportunities to test the

final data dashboard extensively in the environment(s) in which the dashboard

will ultimately be used to ensure the dashboard is user friendly, functions in the
manner expected by its users, and produces the outputs expected and needed by
its users.
348 Chapter 8 Data Dashboards

In addition, the design team must work with the management team and the users to create
a process for ongoing monitoring and revision of the data dashboard to ensure it continues
to effectively meet the organization’s needs. Following a process that adheres to these
guidelines ensures the data dashboard will generate value for the organization throughout
its life.

N otes 1 C omments

1. One important limitation of Excel PivotTables is that they 4. Excel’s Timeline works in the same way a slicer works, but
can only be used to create column, bar, line, pie, and radar the timeline works exclusively with date fields to provide
PivotCharts. a way to filter and group the dates in a PivotTable. The
2. Slicers filter PivotTables, so data tables with user control Timeline button can be found adjacent to the Slicer button
can be included on a data dashboard if that will be useful in the Filters group on the Insert tab.
to the users. 5. Excel’s Developer Tab is another set of tools that can be
3. Borders can be removed from each slicer by clicking the used to allow the user to interact with a data dashboard.
slicer, clicking the Slicer Tools Options tab on the Rib- The functionality of several of the Developer Tab tools can
bon, right clicking the current style in Slicer Styles group, be incorporated into dashboards through slicers, and other
clicking Duplicate, selecting Whole Slicer and clicking the Developer Tab tools require macros to be written. The
Format button in the Modify Slicer Style dialog box, then Developer Tab is generally hidden and must be activated
clicking the Border tab and selecting None. This creates in order to be added to the Ribbon.
a new style for Slicer Styles in the Slicer Tools Options 6. A variety of commercial products for developing data dash-
tab on the Ribbon that you can apply to any slicer in your boards are available. Products such as Tableau, Domo,
workbook. Qrvey, GROW, Microsoft Power BI, and ClicData can be
used to create complex data dashboards.

8-5 Common Mistakes in Data Dashboard

Design
Although there are many types of mistakes one can make in designing and developing a
data dashboard, a few of these types of mistakes are common. These include:
●● Not considering the reason the organization wants to develop a data dashboard,
i.e., the organization’s needs and objectives. This includes
❯❯ neglecting information that is important with respect to the objectives of the
organization
❯❯ focusing on information that is not meaningful with respect to the objectives
of the organization
●● Neglecting to obtain sufficient input from the actual users throughout the data dash-
board design process
●● Not considering the environment(s) in which the data dashboard will be used

●● Failing to position complimentary components (charts, tables, etc.) on the data

dashboard in a manner that facilitates their simultaneous use

●● Using an inappropriate or ineffective type of chart for the data and its message

●● Neglecting the principles of good chart and table design when creating the individual

components of the data dashboard

●● Creating a data dashboard that is too cluttered

●● Designing an unattractive visual display

●● Not considering the organization’s and users’ potential future needs

Glossary 349

S U M M A RY

In this chapter, we have discussed how data dashboards can help users understand and
investigate data and ultimately make better decisions. The process of developing a data
dashboard starts with understanding the principles of effective data dashboards, common
applications of data dashboards, and various types of data dashboards.
We reviewed various aspects of data dashboard design. We discussed the need to under-
stand the purpose of the data dashboard and consider the needs of the data dashboard’s users.
We explained the importance of considering the information to be displayed and how it is to be
displayed in the data dashboard. We showed how to use PivotTables, PivotCharts, and slicers to
build a data dashboard in Excel. We discussed how to link slicers to multiple PivotTables and
how to protect a data dashboard so users cannot make permanent changes. We emphasized the
importance of a final review of a data dashboard and consideration of future needs. We con-
cluded by listing several mistakes commonly made in data dashboard design and development.

G L O S S A RY

Analytical dashboard A dashboard typically used by analysts to identify and investigate

trends, predict outcomes, and discover insights in large volumes of data. Because analytical
dashboards usually support exploration of long-term issues, these dashboards are generally
updated less frequently than operational, strategic, or tactical dashboards.
Customization tool A feature that allows the user to tailor the dashboard to specific needs.
Data dashboard A data visualization tool that gives multiple outputs and may update in
real time.
Drilling down A feature that provides the user with more specific and detailed
information on a particular element, variable, or KPI.
Dynamic dashboard A dashboard that automatically receives and incorporates new data
into the dashboard as the new data become available.
Hierarchical filtering A feature that provides the user with the capability to restrict the
data displayed to a specific segment by systematically selecting values of several categories
or values of variables in a nested manner.
Interactive dashboard A dashboard that allows users to customize the data dashboard
display, effectively allowing a user to filter the data displayed on the dashboard.
Key performance indicator A value a manager uses to operate and maintain their
businesses effectively and efficiently. Also known as a KPI.
Noninteractive dashboard A dashboard that does not allow users to customize the data
dashboard display.
Operational dashboard A dashboard typically used by lower level managers to monitor
rapidly changing critical business conditions.
PivotChart A chart that allows the user to interact with the data by applying filters to
select various aspects of the data to be displayed in the chart.
PivotTable A table that allows the user to interact with the data by applying filters to
select various aspects of the data to be displayed in the table.
Slicer A tool that allows the spreadsheet user to filter the data to be displayed in
PivotTables and PivotCharts.
Static dashboard A dashboard that may periodically be updated manually as new data
and information are collected.
Strategic dashboard A dashboard typically used by executives to monitor the status of
KPIs relevant to overarching organizational objectives.
Tactical dashboard A dashboard typically used by mid-level managers to identify and
assess the organization’s strengths and weaknesses in support of the development of
organizational strategies.
Time interval widget A feature that allows the user to specify the time period to be
displayed on a data dashboard.
350 Chapter 8 Data Dashboards

P R O B L E M S

Conceptual
1. Definition of a Data Dashboard. Which of the following is true with regard to a data
dashboard? LO 1
i. It is a data visualization tool that gives multiple outputs and may update in real
time.
ii. Its outputs are often a set of KPIs for the company or some unit of the company
that are aligned with organizational goals.
iii. It can help an organization better understand and use its data.
iv. All of the above are true with regard to a data dashboard.
2. Understanding the Objectives of Effective Data Dashboards. Ideally, a data
dashboard for an organization should do which of the following? LO 2
i. Present all KPIs to provide the broadest information possible.
ii. Present unrelated KPIs to provide the most contrasting information possible.
iii. Present both unrelated and related KPIs to provide the broadest information
possible.
iv. Present KPIs related to some aspect of the organization’s operations to provide
information relevant to a specific problem or concern.
3. Understanding Principles of Effective Data Dashboards. Which of the following is
true when designing a data dashboard? LO 2
i. one does not need to be concerned with adhering to the principles of effective
data visualization for the dashboard or any of its components (charts, tables, etc.).
ii. one should adhere to the principles of effective data visualization for all com-
ponents (charts, tables, etc.) of the dashboard but not necessarily for the overall
dashboard.
iii. one should adhere to the principles of effective data visualization for the overall
dashboard but not for any of the components (charts, tables, etc.) of the dash-
board but not necessarily for the overall dashboard.
iv. one should adhere to the principles of effective data visualization for the dash-
board and each of its components (charts, tables, etc.).
4. Common Areas of Application for Data Dashboards. Which type of dashboard is
used to monitor performance of an organization’s workforce by individual, division or
department, or shift? LO 3
i. Technical support
ii. Marketing
iii. Human resource
iv. Investment
5. Common Areas of Application for Data Dashboards. Which type of dashboard is
used to monitor contributor activity and potential for nonprofit organizations by con-
tributor, prospective contributor, or project/campaign? LO 3
i. Technical support
ii. Donor
iii. Marketing
iv. Manufacturing
6. Data Dashboard Taxonomies. Which type of dashboard might provide information
on a user’s pulse, blood pressure, weight, body-mass index, and calories consumed in a
day? LO 4
i. Technical support
ii. Manufacturing
iii. Investment analytics
iv. Personal fitness
Problems 351

7. Data Dashboard Taxonomies. Which of the following most accurately describes a

dynamic dashboard? LO 4
i. It regularly receives and incorporates new and revised data and incorporates these
data into the dashboard.
ii. It provides information in a lively manner.
iii. It incorporates animation.
iv. It is most useful when the organization’s KPIs change slowly.
8. Characteristics of an Interactive Data Dashboard. Which of the following most
accurately describes an interactive dashboard? LO 4
i. It regularly receives and incorporates new and revised data and incorporates these
data into the dashboard.
ii. It allows a user to permanently alter the data displayed on the dashboard.
iii. It allows a user to filter the data displayed to her or him on the dashboard.
iv. It incorporates audio responses to the user.
9. Restricting User Interaction in a Data Dashboard. Which of the following features
allow the user to restrict the data displayed to a specific segment of users by systemati-
cally selecting values of several categories or values of variables in a nested manner? LO 4
i. Hierarchical filtering
ii. Data cleansing
iii. Information sifting
iv. Methodical selection
10. Data Dashboard Taxonomies. Identify which type of data dashboard is best described
by each of the following descriptions. LO 4
a. Dashboards typically used by executives to monitor the status of KPIs relevant
overarching organizational objectives.
i. Operational
ii. Tactical
iii. Strategic
iv. Analytic
b. Dashboards typically used by analysts to identify and investigate trends, predict
outcomes, and discover insights in large volumes of data.
i. Operational
ii. Tactical
iii. Strategic
iv. Analytic
c. Dashboards typically used by mid-level managers to identify and assess the organi-
zation’s strengths and weaknesses in support of the development of organizational
strategies.
i. Operational
ii. Tactical
iii. Strategic
iv. Analytic
d. Dashboards typically used by lower level managers to monitor rapidly changing
critical business conditions.
i. Operational
ii. Tactical
iii. Strategic
iv. Analytic
11. Overcrowding and Unnecessary Complexity in a Data Dashboard. Which of the fol-
lowing is not a valid strategy for avoiding overcrowding and unnecessary complexity in
data dashboard design? LO 5
i. Using as many different colors as possible on the charts and tables used in the
dashboard
352 Chapter 8 Data Dashboards

ii. Using interactive dashboard tools (drilling down, hierarchical filtering, time inter-
val widget, customization tools)
iii. Organizing the information into subsets that each address a different need
of the end users and displaying information in these subsets across multiple
pages
iv. Avoiding inclusion of information that will not be useful to the end users
12. Providing Effective Context in a Data Dashboard. Which of the following is an effec-
tive way to provide context for the information provided by the data dashboard? LO 5
i. Comparing the value of a KPI to an organizational goal
ii. Showing how a KPI varies over time
iii. Comparing the value of a KPI across customers
iv. Each of the above is an effective way to provide context for the information pro-
vided by the data dashboard.
13. The Data Dashboard Testing Process. Which of the following is not an important
part of the process of testing a data dashboard throughout its development? LO 5
i. Reviewing progress with the management team that authorized the development
of the data dashboard at critical junctures to ensure the dashboard under produc-
tion is meeting the management team’s objectives
ii. Allowing the general public to use and comment on the data dashboard at various
stages of development to ensure the dashboard can be used by anyone
iii. Allowing the ultimate users to test and comment on the data dashboard at various
stages of development to ensure the dashboard is user friendly, functions in the
manner expected by the users, and produces the outputs expected and needed by
its users
iv. Testing of the dashboard at each step by the dashboard design team to ensure the
dashboard functions as the team intends
14. Excel Tools for Filtering Data in a Data Dashboard. Which of the following is an
Excel tool that allows the dashboard user to filter the data to be displayed in
PivotTables and PivotChart LO 6
i. Screener
ii. Dicer
iii. Slicer
iv. Strainer
15. Excel Tables and Building Data Dashboards. What is the advantage of creating an
Excel Table from the raw data in Excel? LO 6
i. You can add a new record to the existing data by entering the record into the row
adjacent to the last row of the table.
ii. You can add a new field to the existing data by entering the field into the column
adjacent to the last column of the table.
iii. You can give the table a name and refer to the table by that name instead of its
range of cells.
iv. Each of the above is an advantage of creating a table from the raw data in Excel.
16. Common Mistakes in Data Dashboard Design. Which of the following is a com-
mon mistake in designing and developing a data dashboard? LO 7
i. Using copious amounts of animation to engage and entertain the user
ii. Using a different type of chart for each display in a data dashboard to provide the
user with visual variety
iii. Not giving careful consideration to the environment(s) in which the data dash-
board will be used
iv. Randomly providing unrelated but interesting trivia about the organization at the
bottom of the data dashboard to encourage the user to return often
Problems 353

Applications

17. Evaluating the Design of a Data Dashboard. An alternative version of the Esplén-
dido Jugo y Batido data dashboard shown in Figure 8.12 follows. How would you
modify this alternative to improve the dashboard? LO 2

18. Evaluating the Choice of Charts in a Data Dashboard. An alternative version of the
Espléndido Jugo y Batido data dashboard shown in Figure 8.12 follows. How would
you modify this alternative to improve the dashboard? LO 2

19. Adding a Chart to a Data Dashboard. In addition to the information provided on

the data dashboard shown in Figure 8.12, Espléndido Jugo y Batido management has
decided it would like to include a series of charts that show the relationship between
average service satisfaction rating and average product satisfaction rating across the
distribution centers for each of the past three years. EJB Management would also like
to include a table that provides average service satisfaction rating and average product
satisfaction rating by distribution center across years. LO 2
a. What type of chart(s) would you recommend to show the relationship between the
average of the service satisfaction rating and the average of product satisfaction
rating across the distribution centers for each of the past three years? explain why
you would suggest such chart(s).
b. How would you revise the data dashboard in Figure 8.12 to incorporate the chart or
charts you suggested in part a?
354 Chapter 8 Data Dashboards

20. Building a Technical Support Data Dashboard. Bogdan’s Express, a chain of sport-
ing goods stores in Washington, wants to construct a technical support data dashboard
to monitor how effectively its technical support group deals with IT problems. Man-
agement is primarily interested in the time it takes the IT support group to respond
once a problem has been reported (response time) and how long it takes the group to
resolve the issue after the initial response by the IT support group (time to resolution)
over the most recent four months. They would like to be able to review the IT group’s
performance by date, type of technical problem (email, hardware, or internet), and
office (Bellingham, Olympia, Seattle, or Spokane).
Each reported problem is immediately logged and issued a case number, and the data
collected by Bogdan from its relational database includes the case number, date, office,
type of technical problem, response time (in minutes), and time to resolution (in min-
utes). They have also created a new field for the month during which the problem was
reported. Note that both Response Time and Time to Resolution only include time that
elapses during normal business hours.
Bogdan’s staff has already created the following components for the data dashboard
in Excel. LO 6
●● A line chart of average response time across months by office (in the Chart1 worksheet)

●● A line chart of average time to resolution across months by office (in the Chart2

worksheet)
●● A clustered bar chart of average time to resolution across offices by type of technical

problem (in the Chart3 worksheet)

●● A stacked cluster column chart of number of problems reported across office by type

of technical problem (in the Chart4 worksheet)

The data and these charts are available in the file BogdanCharts, and the name of the
worksheet in which each chart can be found is given with the descriptions of the charts
Bogdan has developed.
BogdanCharts
a. Create the following slicers for the charts created by Bogdan.
●● A slicer for month and a slicer for office for the chart in the Chart1 worksheet

●● A slicer for month and a slicer for office for the chart in the Chart2 worksheet

●● A slicer for office and a slicer for type of technical problem for the chart in the

Chart3 worksheet
●● A slicer for office and a slicer for type of technical problem for the chart in the

Chart4 worksheet
b. Create the data dashboard by creating a new worksheet and naming it Dashboard;
moving the charts and slicers from the Chart1, Chart2, Chart3, and Chart4 work-
sheets to the Dashboard worksheet; repositioning these charts and slicers on the
Dashboard worksheet; and adding a title to the dashboard and doing whatever
formatting and editing is necessary to make the dashboard functional and visually
appealing.
c. Amend the data dashboard you created in part b in the following ways.
●● Create a single slicer to filter month for the charts created in the Chart1 and

Chart2 worksheets
●● Create a single slicer to filter office for the charts created in the Chart1, Chart2,

Chart3, and Chart4 worksheets

●● Create a slicer to filter type of technical problem for the charts created in the

Chart3 and Chart4 worksheets

Once you have amended the data dashboard, rearrange the charts and slicers to cre-
ate an effective and visually appealing dashboard. Test each slicer to ensure it works
on the appropriate charts.
d. Protect the data dashboard in part c from being revised by users. Ensure that the slic-
ers cannot be resized or moved, password protect the Dashboard worksheet (use the
password Problem820), and hide all worksheets except the Dashboard worksheet.
 Problems 355

e. The following seven entries for April 29–30 in the following table were not logged. Add
these data to the BogdanData table and refresh all PivotTables and PivotCharts. Com-
ment on differences between the resulting dashboard and the dashboard from part d.

Type of Response Time to

Case Technical Time Resolution
Number Date Month Office Problem (Minutes) (Minutes)
1990 29-Apr Apr Olympia hardware 59.9 23.1
1994 29-Apr Apr Spokane hardware 15.8 64.1
2000 29-Apr Apr Bellingham hardware 26.7 53.4
2005 29-Apr Apr Bellingham internet 41.4 12.8
2011 30-Apr Apr Olympia hardware 96.1 14.6
2012 30-Apr Apr Spokane hardware 125.6 55.0
2019 30-Apr Apr Spokane email 4.5 45.9

21. Building a Donor Data Dashboard. The American Retriever Foundation (ARF) is a
not-for-profit organization dedicated to health issues faced by the six distinct retriever
breeds (Chesapeake Bay, curly-coated, flat-coated, golden, Labrador, and Nova Scotia
duck tolling). ARF needs to develop a data dashboard to monitor its donor activity and
its interactions with potential donors. Management is concerned primarily with the
number and dollar value of donations, the number of legacy donors (those who have
donated in the past twelve months) and new potential donors solicited, and the num-
ber of solicitations that result in donations. They want to compare these results across
ARF’s four development officers (Randall Shalley, Donna Sanchez, Marie Lydon, and
Hoa Nguyen) by date and mode of contact (telephone, email, or personal meeting).
ARF has collected data for each solicitation initiated last year from its relational
database. These data include the solicitation number, development officer, date of
solicitation, mode of solicitation, whether the solicitation resulted in a donation, and
whether the solicited potential donor was a legacy donor. ARF also added a field for
month in which the solicitation was made.
ARF’s staff has already created the following components for the data dashboard in Excel.
●● A stacked bar chart of the number of solicitations across development officer by
whether the solicitation resulted in a donation (in the Chart1 worksheet)
●● A stacked bar chart of the percentage of successful solicitations across mode of so-

licitation by whether the solicitation resulted in a donation (in the Chart2 worksheet)
●● A stacked bar chart of total donations (in $1000s) across development officer by

legacy designation (in the Chart3 worksheet)

●● A cluster bar chart of total donations (in $1000s) across mode of solicitation by

legacy designation (in the Chart4 worksheet)

The data and these charts are available in the file ARFCharts, and the name of the
worksheet in which each chart can be found is given with the descriptions of the charts
ARF has developed. LO 6
ARFCharts
a. Create the following for the charts created by ARF.
●● A slicer for development officer, a slicer for whether the solicitation resulted

in a donation, and a slicer for month of solicitation for the chart in the Chart1
worksheet
●● A slicer for mode of solicitation, a slicer for whether the solicitation resulted in a

donation, and a slicer for development officer for the chart in the Chart2 work-
sheet
●● A slicer for development officer, a slicer for legacy status of the donor, and a

slicer for month of solicitation for the chart in the Chart3 worksheet
●● A slicer for mode of solicitation, legacy status, and month of donation for the

chart in the Chart4 worksheet

356 Chapter 8 Data Dashboards

b. Create a data dashboard by creating a new worksheet and naming it Dashboard;

moving the charts and slicers from the Chart1, Chart2, Chart3, and Chart4
worksheets to the Dashboard worksheet; repositioning the charts and slicers
on the Dashboard worksheet; and adding a title to the dashboard and doing
whatever formatting and editing is necessary to make the dashboard functional
and visually appealing.
c. Amend the data dashboard you created in part b in the following ways.
●● Create a single slicer to filter month for the charts created in the Chart1, Chart2,

and Chart4 worksheets

●● Create a single slicer to filter development officer for the charts created in the

Chart1, Chart2, and Chart3 worksheets

●● Create a single slicer to filter whether or not a solicitation resulted in a donation

for the charts created in the Chart1 and Chart2 worksheets

●● Create a single slicer to filter legacy status for the charts created in the Chart3

and Chart4 worksheets

●● Create a single slicer to filter mode of solicitation for the charts created in the

Chart2 and Chart4 worksheets

Once you have amended the data dashboard, rearrange the charts and slicers to cre-
ate an effective and visually appealing dashboard. Test each slicer to ensure it works
on the appropriate charts.
d. Protect the data dashboard in part c from being revised by users. Ensure that the slic-
ers cannot be resized or moved, password protect the Dashboard worksheet (use the
password Problem821), and hide all worksheets except the Dashboard worksheet.
e. The file ARFNewData contains 15 entries for the past year that were not originally
entered into the relational data base before the data for this dashboard was retrieved.
Add these data to the ARFData table and refresh all pivot tables and pivot charts.
ARFNewData
Comment on differences between the resulting dashboard and the dashboard from
part d.
22. Building a Personal Fitness Data Dashboard. Late last year, Veronica Camp learned
she is pre-diabetic. Her physician recommended that she immediately begin exercising
regularly and limiting her calorie intact to approximately 1500 calories per day. Her
physician also advised her to monitor her systolic and diastolic blood pressure, heart
rate, and blood glucose daily.
On January 1, Veronica began exercising three times a week for approximately
30 minutes, and she started monitoring her calorie intake at meals with an app she
downloaded. She also started monitoring her systolic and diastolic blood pressure,
heart rate, and blood glucose daily. She has created a database with these data recorded
on a daily basis through November, and she has generated the following charts. LO 6
●● A line chart of monthly average systolic blood pressure (in the Chart1 worksheet)

●● A line chart of monthly average diastolic blood pressure (in the Chart2 worksheet)

●● A line chart of monthly average heart rate (in the Chart3 worksheet)
VeronicaCharts
●● A line chart of monthly average blood glucose level (in the Chart4 worksheet)

●● A line chart of monthly total minutes of exercise (in the Chart5 worksheet)

●● A line chart of average daily calorie intake for each meal across months (in the

Chart6 worksheet)
a. Create a data dashboard for Veronica by creating a new worksheet and naming it Dash-
board; moving the charts from the Chart1, Chart2, Chart3, Chart4, Chart5, and Chart6
worksheets to the Dashboard worksheet; and repositioning the charts on the Dashboard
worksheet, adding a title to the dashboard, and doing whatever formatting and editing is
necessary to make the dashboard functional and visually appealing.
b. Create a single slicer in the data dashboard you created in part a that filters day of
the week for all charts on the dashboard. Once you have amended the data dash-
board, rearrange the charts and slicers to create an effective and visually appealing
dashboard. Test the slicer to ensure it works on the appropriate charts.
 Problems 357

c. Create the new field Total Calories in column L of the VeronicaData Excel Table by
summing Breakfast Calories, Lunch Calories, Dinner Calories, and Dessert Calo-
ries for each day. Refresh all PivotTables and PivotCharts, then add average daily
total calories to the line chart of average daily calorie intake for each meal across
months. Comment on differences between the resulting dashboard and the dash-
board from part b.
d. Protect the data dashboard in part c from being revised by users. Ensure that the slicer
cannot be resized or moved, password protect the Dashboard worksheet (use the
password Problem822), and hide all worksheets except the Dashboard worksheet.
23. Building a Baseball Statistics Data Dashboard. The Springfield Spiders, a baseball
team in the All American Baseball Association, wants to create a data dashboard for its
fans. Spiders management would like the fans to be able to review the runs scored and
allowed by game, and review the number of wins and losses and the average per game
attendance by opponent and by day of the week. They would also like for the fans to be
able to filter each of these displays by home and away games.
The Spiders have collected data on the date, opponent, whether the game was played
at home or away, how many runs the Spiders scored, how many runs the Spiders
allowed their opponent to score, whether the Spiders won or lost, and the attendance
for each game of the previous season. They also added a field for the day of the week,
and have created the following charts for inclusion in its data dashboard. LO 6
●● A line chart of runs scored and runs allowed by game (in the Chart1 worksheet)

●● A clustered column chart of number of wins and losses by month (in the Chart2

worksheet)
SpidersCharts
●● A clustered bar chart of average per game attendance across months for home and

away games (in the Chart3 worksheet)

●● A clustered bar chart of average per game attendance across days of the week for

home and away games (in the Chart4 worksheet)

a. Create a data dashboard for the Spiders by creating a new worksheet and naming it
Dashboard; moving the charts from the Chart1, Chart2, Chart3, and Chart4 work-
sheets to the Dashboard worksheet; and repositioning the charts on the Dashboard
worksheet, adding a title to the dashboard, and doing whatever formatting and
editing is necessary to make the dashboard functional and visually appealing.
b. Amend the data dashboard you created in part b in the following ways.
●● Create a slicer to filter day of the week for the charts created in the Chart1,

Chart2, Chart3, and Chart4 worksheets

●● Create a slicer to filter opponent for the charts created in the Chart1, Chart2,

Chart3, and Chart4 worksheets

●● Create a slicer to filter month for the charts created in the Chart1, Chart2,

Chart3, and Chart4 worksheets

●● Create a slicer to filter home and away games for the charts created in the Chart1,

Chart2, Chart3, and Chart4 worksheets

Once you have amended the data dashboard, rearrange the charts and slicers to cre-
ate an effective and visually appealing dashboard. Test each slicer to ensure it works
on the appropriate charts.
c. Create an alternative display of the Runs Scored and Runs Allowed by game
through the following steps.
Step 1. Create the new field Run Differential in column I of the SpidersData table
by subtracting Runs Allowed from Runs Scored for each game
Refresh all PivotTables and PivotCharts
Step 2. Create a new worksheet and rename the worksheet Chart1A
Create a new pivot table and pivot chart in the Chart1A worksheet
Drag the Date field to the Rows area and the Run Differential field to
the Values area in the PivotFields Table pane to create a column chart
358 Chapter 8 Data Dashboards

Right click the horizontal axis and click Format Axis... to open the
Format Axis pane, click the Axis Options button and click Labels, then
select Low from the Label Position dropdown menu to position the hori-
zontal axis at the bottom of the chart
While still in the Format Axis task pane, click the Axis Options button
and click Fill & Line, then in the Line area change the color of the axis
to black and increase its width to 2 pt
Click any bar in the chart and click Format Data Series, click the Fill
& Line button, and in the Fill area click the Invert if Negative box and
select the colors to be applied to the positive and negative bars
Step 3. Replace the line chart on the dashboard that was originally created in the
Chart1 worksheet with the chart you created in the Chart1A worksheet
Comment on differences between the resulting dashboard and the dashboard from
part b.
d. Protect the data dashboard in part c from being revised by users. Ensure that the
slicers cannot be resized or moved, password protect the Dashboard worksheet (use
the password Problem823), and hide all worksheets except the Dashboard work-
sheet.
e. Data for the first month of the new season is available in the file SpidersNewData. Add
this new data to the SpidersData table and delete the data for the previous season in
the SpidersData table in the Data worksheet of the dashboard from part c so that the
SpidersNewData
dashboard displays only the new season data. What does the resulting data dashboard
communicate? What alternative approach to incorporating the new data into the exist-
ing data dashboard would you suggest?
Chapter 9
Telling the Truth with Data Visualization
CONTENTS

DATA VISUALIZATION MAKEOVER 9-4 Deceptive Design

Top-10 Movies of All Time Design of Chart Axes
9-1 Missing Data and Data Errors Dual-Axis Charts
Identifying Missing Data Data Selection and Temporal Frequency
Identifying Data Errors Issues Related to Geographic Maps

9-2 Biased Data Summary

Selection Bias Glossary
Survivor Bias
Problems
9-3 Adjusting for Inflation

L E A R N I N G O B J E C T I V ES
After completing this chapter, you will be able to

LO 1 Identify missing data and data errors using Excel. LO 6 Explain why dual-axis charts are often confusing
and misleading for audiences and suggest an
LO 2 Define the meaning of biased data and explain
alternative to using a dual-axis chart that is less
the concepts of selection bias and survivor bias.
confusing for the audience.
LO 3 Define Simpson’s paradox and explain how
LO 7 Explain how the range of data and the temporal
a scatter chart can be used to identify some
frequency of data included in a chart affects the
instances of Simpson’s paradox.
insights conveyed to the audience.
LO 4 Explain the importance of adjusting for inflation
LO 8 Explain why some geographic maps can result in
in time series data that represent long time peri-
misleading data visualizations and provide rec-
ods and use a price index to adjust nominal val-
ommendations for how to improve these types of
ues to account for inflation.
maps.
LO 5 Identify deceptive design practices related to the
axes used in charts and suggest ways to improve
the axes in these charts to communicate insights
to the audience more clearly.
 Data Visualization Makeover 361

D ata V is u ali z ation M a k eover

Top-10 Movies of All Time

Many lists of the top movies of all time can be found top 10 is Titanic, which was released in 1997. The
on the internet. Some of these lists are subjective and nine other movies shown here were released since
based on movie ratings or opinions. Other lists are 2009.
based on quantitative data such as revenues earned, Based on Figure 9.1, we might conclude that the
tickets sold, or profit generated. Box Office Mojo is a most recent years have produced all of the top movies
website that collects data on box office revenue and in terms of greatest box office revenue in North Amer-
makes these data available to the public. Box Office ica. From this figure, it appears that movies prior to
Mojo is owned by the Amazon.com subsidiary IMDb. 1997 were not as popular at the box office as movies
The website boxofficemojo.com contains a great in recent years.
deal of data for movies and box office revenue. The In examining the design of Figure 9.1, we see
website continuously tracks the top movies based on few obvious errors. The chart has an explanatory
box office revenue, and it has collected data for many title, the axes are well labeled, there is no obvious
movies made in the United States over the last 100 evidence of clutter, and the chart is relatively simple
years. with a high data-ink ratio. However, we still must
Figure 9.1 shows the all-time top-10 movies take great care in interpreting this chart. A common
made in the United States based on box office reve- problem arises when examining monetary units such
nues earned by movies in North American theaters. as revenues, costs, or profits at different points of
In Figure 9.1, the horizontal axis shows the year time. This is because $500,000 earned in box office
the movie was released and the vertical axis shows revenue in 1951 is different than $500,000 earned in
the total box office gross revenue earned in North box office revenue in 2020. This difference is due to
America. Interestingly, we see that all of the top-10 inflation. According to IMDb,1 the average price of
movies were made recently; the oldest movie in the a movie ticket in 1951 was $0.53, while the average

Figure 9.1 Scatter Chart Showing the Top-10 Movies of all Time Based on Lifetime
North America Box Office Gross Revenues
Top-10 Movies of All Time
Lifetime North America Box Office Gross Revenue ($ millions)
1,000
Star Wars: Episode VII
900 - The Force Awakens
Avengers: Endgame
800 Black
Avatar
Panther
700 Avengers: Infinity War
Titanic
600 Jurassic World

Incredibles2
500 The Avengers
Star Wars:
400 Episode VIII
- The Last
300 Jedi

200
100
0
1930 1940 1950 1960 1970 1980 1990 2000 2010 2020
Year of Release

1
See https://help.imdb.com/article/imdbpro/industry-research/box-office-mojo-by-imdbpro-faq
/GCWTV4MQKGWRAUAP?ref_=mojo_cso_md#inflation.

(Continued)
362 Chapter 9 Telling the Truth with Data Visualization

Figure 9.2 Scatter Chart Showing the Top-10 Movies of All Time Based on Inflation-Ad-
justed Lifetime North America Box Office Gross Revenue

Top-10 Movies of All Time

Lifetime North America Box Office Gross Revenue, Adjusted for Inflation ($ millions)
2,000 Gone with the Wind
Star Wars: Episode IV - A
1,800 New Hope

1,600
E.T. the Extra-
The Sound of Music
1,400 Terrestrial
Titanic
The Ten Commandments Jaws
Doctor Zhivago
1,200 Snow White and the Seven
Dwarfs The Exorcist
1,000

800

600

400

200

0
1930 1940 1950 1960 1970 1980 1990 2000 2010 2020
Year of Release

price of a movie ticket in 2020 was $9.37. Therefore, inflation. Gone with the Wind, which was released
to earn $500,000 in revenue in 1951, a movie would in 1939, actually generated the all-time highest box
need to sell $500,000 / $0.53 5 943,396 tickets at office gross revenue in North America when we adjust
the box office. To earn $500,000 in revenue in 2020, for inflation. We also see from Figure 9.2 that no
a movie would only need to sell $500,000 / $9.37 5 movie released after the year 2000 is in the top 10
53,362 tickets. In other words, selling 5.6% as many of box office gross revenue after adjusting for infla-
tickets for a movie in 2020 as a movie in 1951 gen- tion. Comparing Figure 9.1 to Figure 9.2, we see that
erates the same amount of box office revenue. This only one movie, Titanic, which was released in 1997,
means it is difficult for older movies to appear on appears in both figures.
this chart of top-10 movies because older movies Figures 9.1 and 9.2 illustrate the importance of
would have to sell many more tickets at the box adjusting for inflation when visualizing monetary
office than more recent movies. This effectively values that occurred at different points of time. It is
biases our data to more heavily weight box office generally not important to adjust for inflation when
ticket sales of newer movies. all monetary values are from similar time periods
To help remove the bias in these data, we can (usually within a few years), but failing to adjust
measure the top-10 movies of all time based on for inflation when monetary values occur across
inflation-adjusted box office gross revenue. Later wide spans of time can produce extremely biased
in this chapter we will explain in more detail how visualizations. The insight provided by Figure 9.1 is
to adjust for inflation, but for now we explain the that all of the highest box office revenue producing
concept as simply that we measure the box office movies were released fairly recently, with only one
gross revenue in terms of 2020 dollars. This allows movie in the top 10 being released prior to the
us to compare movies released in different years as year 2000. However, by showing values adjusted for
if all the box office revenue was earned in the year inflation, Figure 9.2 provides a drastically different
2020. Figure 9.2 shows the top-10 movies based on insight—all of the top-10 highest box office revenue
inflation-adjusted box office gross revenue. producing memories were released prior to the year
From Figure 9.2, we see that many older movies 2000.
generated some of the highest values of box office It is also important to note that inflation is only
gross revenue in North America after adjusting for one potential cause of bias for these data. When
 9-1 Missing Data and Data Errors 363

comparing data from different points in time, there were in 1951. This means that there is more com-
are many other factors that could affect the data. petition, and more movies competing to be shown
For instance, the population of North America has in the theaters. Until the 1950s, movies also did not
increased substantially between 1951 and 2020. As have meaningful competition for an audience from
the population has increased, so has the number of television. Adjusting for inflation removes one poten-
movie theaters. This means that there are many more tially substantial source of bias in the data, but it is
movie theaters in existence in 2020 than in 1951. important to realize that there can be other sources
Thus, more recent movies can be shown in more of bias. The designer of a data visualization needs to
theaters and possibly sell more tickets to generate consider the impact these potential biases may have
more box office revenue. However, there are also on their visualization and what steps should be taken
many more movies being released in 2020 than there to mitigate these biases.

In this chapter we discuss how to design data visualizations to tell the truth and how
to avoid creating misleading or confusing visualizations. Some data visualizations do
not tell the complete truth because the designer is actively trying to mislead or unduly
influence the audience. More often, data visualizations do not tell the complete truth
because of unintentional mistakes made by the designer that mislead or cause confu-
sion for the audience. As shown in the Data Visualization Makeover for this chapter,
considerations as simple as not adjusting for inflation can completely change the
resulting data visualization and the insights drawn from the visualization. The issues
covered in this chapter will help create data visualizations that are clear and truthful
for the audience.
All data visualizations begin with data. If we use wrong, incomplete, or biased data,
then our visualization will not tell the complete truth. Therefore, we begin this chapter
by talking about common issues with data, such as missing data and biased data, that can
cause misleading or confusing data visualizations. We discuss deceptive data visualization
designs that should be avoided when creating visualizations. These deceptive designs often
involve improper choices related to the design of chart axes or choices on which data to
include in the visualization. We also discuss issues that can arise when creating geographic
maps, and we provide recommendations for how to avoid misleading charts when creating
these types of maps.

9-1 Missing Data and Data Errors

Many real-life data sets suffer from missing data values and data errors. There are many
reasons why data sets may contain missing values and data errors, and it is important to
understand the reasons behind the missing values in order to know how they should be
treated. Removing or replacing missing data values and errors in data affects all subsequent
analysis of that data, including the creation of data visualizations. In this section, we exam-
Types of missing data are ine some common methods of identifying missing data and data errors, and we discuss
discussed in Chapter 6. methods for dealing with missing data and data errors.

Identifying Missing Data

Consider the case of Blakely Tires, a producer of automobile tires located in the United
States. In an attempt to learn about the conditions of its tires on automobiles in the state
of Texas, the company has obtained information for each of the four tires from 116
automobiles with Blakely brand tires that have been collected through recent state auto-
mobile inspection facilities in Texas. The data obtained by Blakely includes the position
of the tire on the automobile (left front, left rear, right front, right rear), age of the tire,
364 Chapter 9 Telling the Truth with Data Visualization

mileage on the tire, and depth of the remaining tread on the tire. Before Blakely man-
agement attempts to learn more about its tires on automobiles in Texas, it wants to assess
the quality of these data. The first few rows of the data collected by Blakely are shown in
Figure 9.3.

FIGURE 9.3 Portion of Excel Spreadsheet Showing Blakely Tires

Data

The tread depth of a tire is a vertical measurement between the top of the tread rub-
ber to the bottom of the tire’s deepest grooves, and it is measured in 32nds of an inch
in the United States. New Blakely brand tires have a tread depth of 10/32nds of an
inch, and a tire’s tread depth is considered insufficient if it is 2/32nds of an inch or less.
Shallow tread depth is dangerous as it results in poor traction and so makes steering the
automobile more difficult. Blakely’s tires generally last for four to five years or 40,000
to 60,000 miles.
We begin assessing the quality of these data by determining which (if any) observations
have missing values for any of the variables in the Blakely Tires data. We can do so using
Excel’s COUNTBLANK function. The following steps show how to count the missing
observations for each variable in the file BlakelyTires.

Step 1. Enter the heading # of Missing Values in cell G2

Step 2. Enter the heading Life of Tire (Months) in cell H1, the heading Tread Depth in
cell I1, and the heading Miles in cell J1
BlakelyTires
Step 3. Enter the formula 5COUNTBLANK(C2:C457) in cell H2
Copy cell H2 to cells I2 and J2

The result in cell H2 shows that none of the observations in these data is missing its value
for Life of Tire. By repeating this process for the remaining quantitative variables in the
data (Tread Depth and Miles) in columns I and J, we determine that there are no missing
values for Tread Depth and one missing value for Miles. The first few rows of the resulting
Excel spreadsheet are provided in Figure 9.4.
Next we sort all of Blakely’s data on Miles from smallest to largest value to determine
which observation is missing its value of this variable. Excel’s sort procedure will list all
observations with missing values for the sort variable, Miles, as the last observations in the
sorted data.
 9-1 Missing Data and Data Errors 365

FIGURE 9.4 Portion of Excel Spreadsheet Showing Number of Missing Values for Variables in
Blakely Tires Data

We can also use the Excel Conditional Formatting tool to quickly explore the data and
Chapter 6 provides step-by- use visualization to help us identify any missing values. Figure 9.5 indicates that the value
step directions for how to use of Miles is missing from the left front tire of the automobile with ID Number 3354942.
Excel’s Conditional Formatting Because only one of the 456 observations is missing its value for Miles, we may be able to
tool to find missing values.
salvage this observation by logically determining a reasonable value to substitute for this
missing value. It is sensible to assume that the value of Miles for the left front tire of the
automobile with the ID Number 3354942 would be identical to the value of miles for the
other three tires on this automobile, so we sort all the data on ID number and highlight data
values with ID Number 3354942 to find the four tires that belong to this automobile (see
Figure 9.6).

FIGURE 9.5 Using Excel Conditional Formatting Tool to Highlight

Missing Values in Blakely Tires Data
366 Chapter 9 Telling the Truth with Data Visualization

FIGURE 9.6 Portion of Excel Spreadsheet Showing Blakely Tires

Data Sorted from Lowest to Highest by ID Number with
ID 3354942 Highlighted

Figure 9.6 shows that the value of Miles for the other three tires on the automobile
with the ID Number 3354942 is 33,254, so this may be a reasonable value for the Miles
of the left front tire of the automobile with the ID Number 3354942. However, before
substituting this value for the missing value of the left front tire of the automobile
with ID Number 3354942, we should attempt to ascertain (if possible) that this value
is valid—there are legitimate reasons why a driver might replace a single tire. In this
instance we will assume that the correct value of Miles for the left front tire on the auto-
mobile with the ID Number 3354942 is 33,254 and substitute that number in the appro-
priate cell of the spreadsheet.

Identifying Data Errors

Many data sets include errors. These errors could be introduced because of manual entry
mistakes, miscalibration of sensors that automatically collect data, or a variety of other
reasons. Examining the variables in the data set by use of summary statistics, frequency
distributions, histograms, scatter charts, and other tools can uncover data quality issues
and outliers. For example, finding the minimum or maximum value for Tread Depth in
the Blakely Tires data may reveal unrealistic values—perhaps even negative values—for
Tread Depth, which would indicate a problem for the value of Tread Depth for any such
observation.
It is important to note here that many software, including Excel, ignore missing val-
ues when calculating various summary statistics such as the mean, standard deviation,
minimum, and maximum. However, if missing values in a data set are indicated with
a unique value (such as 9999999), these values may be used by software when calcu-
lating various summary statistics such as the mean, standard deviation, minimum,
and maximum. Both cases can result in misleading values for summary statistics,
which is why many analysts prefer to deal with missing data issues prior to using
summary statistics to attempt to identify erroneous outliers and other erroneous
values in the data.
We again consider the Blakely Tires data. We calculate the mean and standard devi-
ation of each variable (age of the tire, mileage on the tire, and depth of the remaining
tread on the tire) to assess whether values of these variable are reasonable in general.
 9-1 Missing Data and Data Errors 367

Return to the file BlakelyTires and complete the following steps:

Step 1. Enter the heading Mean in cell G3
Enter the formula 5AVERAGE(C2:C457) in cell H3
BlakelyTires Copy cell H3 to cells I3 and J3
Step 2. Enter the heading Standard Deviation in cell G4
Enter the formula 5STDEV.S(C2:C457) in cell H4
Copy cell H4 to cells I4 and J4
The results in cells H3 and H4 show that the mean and standard deviation for the life
of tires are 23.8 months and 31.83 months, respectively. These values appear to be rea-
sonable for the life of tires in months. The mean and standard deviation for tread depth
are 7.64/12ths of an inch and 2.51/12ths of an inch, respectively, and the mean and
standard deviation for miles are 25,834.54 and 24,143.38, respectively. These values
appear to be reasonable for tread depth and miles. The results of this analysis are pro-
vided in Figure 9.7.
While the summary statistics of mean and standard deviation are helpful in providing
an aggregate perspective on the data, we also need to attempt to determine if there are any
erroneous individual values for our three variables. We start by finding the minimum and
maximum values for each variable. Return again to the file BlakelyTires and complete the
following steps:
Step 3. Enter the heading Minimum in cell G5
Enter the formula 5MIN(C2:C457) in cell H5
Copy cell H5 to cells I5 and J5
Step 4. Enter the heading Maximum in cell G6
Enter the formula 5MAX(C2:C457) in cell H6
Copy cell H6 to cells I6 and J6
The results in cells H5 and H6 show that the minimum and maximum values for Life
of Tires (Months) are 1.8 months and 601.0, respectively. The minimum value of life of
tires in months appears to be reasonable, but the maximum (which is equal to slightly
over 50 years) is not a reasonable value for Life of Tires (Months). To identify the auto-
mobile with this extreme value, we sort the entire data set on Life of Tire (Months) from
smallest to largest and scroll to the last few rows of the data.
We see in Figure 9.7 that the observation with Life of Tire (Months) value of 601.0
is the left rear tire from the automobile with ID Number 8696859. Also note that the left
rear tire of the automobile with ID Number 2122934 has a suspiciously high value for
Life of Tire (Months) of 111. Sorting the data by ID Number and scrolling until we find
the four tires from the automobile with ID Number 8696859, we find the value for Life of
Tire (Months) for the other three tires from this automobile is 60.1. This suggests that the
decimal for Life of Tire (Months) for this automobile’s left rear tire value is in the wrong
place. Scrolling to find the four tires from the automobile with ID Number 2122934, we
find the value for Life of Tire (Months) for the other three tires from this automobile is
11.1, which suggests that the decimal for Life of Tire (Months) for this automobile’s left
rear tire value is also misplaced. Both of these erroneous entries can now be corrected.
By repeating this process for the remaining variables in the data (Tread Depth and
Miles) in columns I and J, we determine that the minimum and maximum values for Tread
Depth are 0.0/12ths of an inch and 16.7/12ths of an inch, respectively, and the minimum
and maximum values for Miles are 206.0 and 107,237.0, respectively. Neither the mini-
mum nor the maximum value for Tread Depth is reasonable; a tire with no tread would not
be drivable, and the maximum value for tire depth in the data actually exceeds the tread
depth on new Blakely brand tires. The minimum value for Miles may be reasonable, but
the maximum value is not. A similar investigation should be made into these and other
extreme values to determine if they are in error and, if so, what might be the correct value.
368 Chapter 9 Telling the Truth with Data Visualization

FIGURE 9.7 Portion of Excel Spreadsheet Showing the Blakely Tires Data Sorted on Life of Tires
(Months) from Lowest to Highest Value and with Calculated Summary Statistics

Note that rows 8 to 451 are Not all erroneous values in a data set are extreme; these erroneous values are much
hidden in Figure 9.7. more difficult to find. However, if the variable with suspected erroneous values has a rel-
atively strong relationship with another variable in the data, we can explore the data set
through data visualization tools such as scatter charts to help us identify data errors. Here
we will consider the variables Tread Depth and Miles. Because more miles driven should
lead to less tread depth on an automobile tire, we expect these two variables to have a nega-
tive relationship. A scatter chart will enable us to see whether any of the tires in the data set
have values for Tread Depth and Miles that are counter to this expectation.
The red ellipse in Figure 9.8 shows the region in which the points representing Tread Depth
and Miles would generally be expected to lie on this scatter plot. The points that lie outside of
this ellipse have values for at least one of these variables that is inconsistent with the negative
relationship exhibited by the points inside the ellipse. If we position the cursor over the point
outside the ellipse that corresponds to relatively high values of Miles and Tread Depth, Excel
will generate a pop-up box that shows that the values of Tread Depth and Miles for this point are
9.7 and 104658, respectively. The tire represented by this point has very high Tread Depth for
this many Miles, which suggests that the value of one or both of these two variables for this tire
may be inaccurate and should be investigated. Note that the other two data points outside the red
ellipse in Figure 9.8 represent the previously identified likely data errors for Tread Depth.
Closer examination of outliers and potential erroneous values may reveal an error or a
need for further investigation to determine whether the observation is relevant to the cur-
rent analysis. A conservative approach is to create two different data visualizations, one
with and one without outliers and potentially erroneous values. If the insights being con-
veyed by the two data visualizations are very different, then you should spend additional
time to track down the cause of the outliers.

N otes 1 C omments

1. Excel’s Data Validation tool can be used to control what a 2. Outliers should only be removed after careful consideration
user can enter into a cell, limiting the ability to create data of their cause and on their effect on the insights drawn from
errors from manual entries. Data validation is implemented the data. If an outlier is due to an obvious data error, then it
in Excel by clicking on the Data tab on the Ribbon and then can be removed or replaced by a corrected value. If outliers
clicking the Data Validation icon in the Data Tools that are not due to obvious data errors are removed, it is
group. This opens the Data Validation dialog box that generally recommended that this removal should be noted
allows you to create rules that define what types of inputs in the data visualization or associated documentation so that
are valid, and prevents the user from entering invalid inputs. the audience knows that the outliers have been removed.
9-2 Biased Data 369

FIGURE 9.8 Scatter Chart for the BlakelyTires Data Used to Identify
Possible Data Errors

9-2 Biased Data

Another data issue that can lead to misleading or incorrect data visualizations is bias.
Biased data exist when the sample data is not representative of the population data that are
under study. There are many causes for biased data, but we will discuss some of the more
common causes of biased data in this section as well as ways that data exploration can
identify potentially biased data.

Selection Bias
Selection bias is a common source of bias in data that can lead to misleading data
visualizations and incorrect insights. Selection bias occurs when data are drawn from
a sample that has not been properly randomized to represent the intended population.
Selection bias occurs frequently in many different fields, including political science.
Consider a political polling firm that wants to poll likely voters in the United States to
determine their preference among candidates in an upcoming election. If the polling
firm attempts to contact potential voters using only landline phone numbers, then the
resulting sample is likely to be biased in terms of the age of respondents. Older people
are much more likely to have landline phones in their homes and to answer them when
called by a political polling firm. In this case, selection bias occurred due to how the
370 Chapter 9 Telling the Truth with Data Visualization

sample was collected, resulting in a sample consisting of potential voters likely to be

much older than the voter population. Therefore, the polling firm must be careful about
using this sample to infer insights about candidate preference among the population of
all likely voters.
Consider a related but slightly different challenge faced by a research group in the
Economics Department at Empire State University. This group of researchers is trying
to analyze factors that affect annual incomes in the United States. The researchers have
gathered sample data from 106 respondents that include many factors including the respon-
dent’s age, place of residence, and annual income for the most recent year. A portion of
these data are shown in Figure 9.9.

FIGURE 9.9 Portion of Data Used to Examine Relationship between

Age and Income

AgeIncome

Trendlines are introduced in To check their data, the researchers generate a simple scatter chart to explore the relation-
Chapter 6. ship between age and annual income for this sample data. This scatter chart is shown in Figure
9.10. The researchers also fit a simple linear trendline to these data as shown in Figure 9.10.
Surprisingly, the researchers find that there is a negative relationship between age and annual
income in these sample data. This seems to contradict the expectation that annual incomes rise
as someone becomes older due to more years of work experience and advancing in their career.
However, further investigation indicates that this negative relationship is caused
by a specific form of selection bias. The researchers investigate how the data were
collected, and they find that the data were collected from residents of three different cities
in the United States: San Francisco, California; Dallas, Texas; and Naples, Florida. These
three different cities are geographically distant from each other with San Francisco in the
west, Dallas in the middle, and Naples in the east. However, this does not necessarily make
them representative of the entire population of the United States. Figure 9.11 shows a scat-
ter plot similar to Figure 9.10 for the same data.
9-2 Biased Data 371

FIGURE 9.10 Scatter Chart Depicting the Relationship between Age and
Annual Income as a Negative Trend

Annual Income ($)

200000

180000

160000

140000

120000

100000

80000

60000

40000

20000

0
0 20 40 60 80
Age (years)

In Figure 9.11, we have used color to differentiate among the locations of the respon-
dents. The researchers have also fit simple linear trendlines to the relation between age and
annual income for each of the different cities.
Figure 9.11 shows that the relation for the data within each city is a positive trend:
as age increases, annual income also increases. However, when we aggregate all the
data together, as in Figure 9.10, the relation between age and annual income appears
to be the opposite in that annual income decreases as age increases. This is because
the ages and annual incomes from the respondents in each of the three different cities
are quite distinct. These demographic differences become apparent when we color
the scatter chart by location in Figure 9.11. Respondents in San Francisco tend to be
younger and have higher incomes, respondents in Naples tend to be older with lower
annual incomes, and respondents in Dallas tend to be somewhere in between.
The effect shown in Figures 9.10 and 9.11 is known as Simpson’s paradox.
Simpson’s paradox occurs when a specific trend that appears in subsets of data disap-
pears or reverses when the subsets are aggregated. In this example, there is a positive
trend between age and annual income within the data from each city, but this trend
appears to reverse when we aggregate the data across all three cities. Simpson’s par-
adox is also a form of selection bias because we have chosen a sample (in this case
using only respondents from three cities) that does not represent the population (the
entire United States).
372 Chapter 9 Telling the Truth with Data Visualization

FIGURE 9.11 Scatter Chart Illustrating Simpson’s Paradox Because the

Relationship between Age and Annual Income Is Positive
for Subsets of the Data

Annual Income ($)

200000 San Francisco

Dallas
180000 Naples
Chapter 6 shows how to
create a scatter chart in Excel
using multiple colors to 160000
differentiate data points.
140000

120000

100000

80000

60000

40000

20000

0
0 20 40 60 80
Age (years)

Survivor Bias
Survivor bias is another common source of bias that can lead to misleading data visual-
izations and incorrect insights. Survivor bias occurs when a sample data set consists of a
disproportionately large number of observations corresponding to positive outcomes for a
particular event. It is closely related to selection bias because, similar to selection bias, the
sample data are not representative of the population that is being studied.
One of the earliest examples of survivor bias was identified in World War II. The com-
mon version of this story is that the United States (U.S.) military was experiencing heavy
losses to its aircraft due to antiaircraft fire and enemy fighters. The U.S. military studied
the damage sustained by returning aircraft to identify where planes were most likely to be
damaged. They found that most damage occurred to certain sections of the tail and wings
of the aircraft similar to what is shown in Figure 9.12. One suggested action from this anal-
ysis was to add armor to these sections of the tail and wings to protect these parts of the
aircraft. However, the mathematician Abraham Wald argued that this was the opposite of
the correct course of action. Because the only data being examined were from planes that
had survived the observed damage, it was likely that damage to these areas of the wings
and tail were actually the least harmful to the aircraft. Therefore, armor should be added to
the areas where surviving planes did not show damage as it was likely that damage to these
areas is what may have caused the non-surviving planes to crash and not return.
For a more recent example of survivor bias, consider the case of Professor Raturi, a
business school professor who studies entrepreneurship and risk behaviors. Professor
Raturi hypothesizes that entrepreneurs are more likely to have a greater risk tolerance
than non-entrepreneurs. To study this hypothesis, Professor Raturi collects data on
9-2 Biased Data 373

FIGURE 9.12 Depiction of Common Areas of Damage to Returning

Planes in World War II

= Damage to aircraft

87 entrepreneurs who guided their companies from start-up to initial public offering
(IPO). Professor Raturi measures the risk tolerance for each of these entrepreneurs by
administering a detailed questionnaire that measures their tolerance to take on finan-
cial risk. The questionnaire results in a score for each entrepreneur that ranges from 1
(lowest risk tolerance) to 5 (highest risk tolerance). Professor Raturi gives this same
questionnaire to a control group of 100 randomly selected individuals who are not entre-
preneurs. The results of Professor Raturi’s research are summarized in Figure 9.13.

FIGURE 9.13 Comparing Average Risk Tolerance of Entrepreneurs to a

Control Group of Non-Entrepreneurs

Risk Tolerance Comparison for Entrepreneurs and Non-Entrepreneurs

Average Financial Risk Tolerance (1 = Lowest; 5 = Highest)
5
95% Confidence Interval

1
Entrepreneurs Non-Entrepreneurs
374 Chapter 9 Telling the Truth with Data Visualization

Figure 9.13 clearly shows that the entrepreneurs have a higher average risk toler-
ance than the control group, even when we compare the 95% confidence intervals. So,
can we conclude that entrepreneurs have higher risk tolerance than non-entrepreneurs?
Not necessarily. Our results here suffer from survivor bias. The only data we have
The visualization of confidence are for successful entrepreneurs—those that survived and made it to an IPO. We have
intervals on sample statistics is no data for unsuccessful entrepreneurs—those whose companies failed prior to IPO.
discussed in Chapter 5.
It is possible that entrepreneurs, whether they are successful or not, have higher risk
tolerances than the general public. But because we have no data on unsuccessful entre-
preneurs, we cannot conclude that entrepreneurs have a higher risk tolerance than
non-entrepreneurs.

9-3 Adjusting for Inflation

Many data visualizations use time series data related to monetary units such as revenues,
costs, and profits. When the time interval over which the data are collected becomes long,
prices are increasing rapidly, or we are comparing monetary units from different countries,
it can be important to adjust for inflation to avoid misleading visualizations and incorrect
Time series data are discussed insights. Inflation refers to the general increase in prices over time, and it is measured by
in Chapter 5. tracking the changes in the price of a standard set of products and services referred to as a
price index. Inflation, and hence the price index, varies over time and also by country or
region due to changes in the prevailing economic conditions. There are many popular price
indexes that are tracked by economic organizations and used for adjusting for inflation
including the consumer price index (CPI) and the producer price index (PPI). It is appro-
priate to use the price index that is most closely associated with the type of products being
analyzed.
In the Data Visualization Makeover at the beginning of this chapter, we illustrated
the importance of adjusting for inflation when attempting to identify the top movies
of all time based on box office revenues. In this section, we will present an additional
example to illustrate this point and explain one way to adjust time-series data for infla-
tion. Consider the price of gasoline in the United States over time. Figure 9.14 displays
a simple time series chart showing the price of gasoline between 1978 and 2017. From
this chart, it appears that gasoline prices in the United States have increased substan-
tially between 1978 and 2017. The data used to create this chart are contained in the
file PriceGasoline. Examining these data shows that the price of gasoline increased
from $0.65 per gallon in 1978 to $2.47 per gallon in 2017. However, these prices
represent the nominal values, meaning that the values have not been adjusted for
inflation.
To allow for meaningful comparisons of the prices of gasoline, we need to adjust the
nominal values for inflation. Figure 9.15 shows a portion of the nominal values of the aver-
age price of gasoline between 1978 to 2017 as well as the average annual price index for
each year. To adjust for inflation, we need to calculate the price of gasoline in each year as
if the gasoline was being sold in one common year, known as the base year. Here we will
define the base year to be the year 2017. Therefore, to adjust for inflation for the price of
gasoline in 1978, we use the calculation

Price Index in 2017 211.8

Price of Gas in 1978 3 5 $0.65 3 5 $2.66.
Price Index in 1978 51.9

This result shows us that although the price of gasoline in 1978 was only $0.65 per
gallon, this is equivalent to a price of $2.66 per gallon in 2017. The general formula used
to find the inflation-adjusted value of a good or service in year x from the nominal value
given in year y is the following.
Price Index in Year x
Inflation adjusted values are Inflation-Adjusted Value in Year x dollars 5 Nominal Value in Year y 3
also known as real values. Price Index in Year y
 9-3 Adjusting for Inflation 375

FIGURE 9.14 Non-inflation Adjusted Price per Gallon of Gasoline in the

United States between 1978 and 2017

Nominal Gas Prices Between 1978 and 2017

Price ($ per gallon)
4.00

3.50

3.00

2.50

2.00

1.50

1.00

0.50

0.00
1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Year
Source: Data from https://www.usinflationcalculator.com/gasoline-prices-adjusted-for-inflation/

FIGURE 9.15 Portion of Data that Contains Nominal Prices and Price
Index Values for Gallon of Gasoline in the United States
between 1978 and 2017

PriceGasoline

Note that rows 11 through 39

are hidden.

Figure 9.16 shows the calculations used in Excel to convert the nominal gasoline prices
into the inflation-adjusted gasoline prices using 2017 as the base year. These values are
then used to create Figure 9.17, which shows the nominal values and the inflation-adjusted
values of the price of gasoline between 1978 and 2002. From Figure 9.17, we see that
when we adjust for inflation, the insight from these time-series data is very different than
Figure 9.15. The inflation-adjusted price of gasoline in the United States has not increased
between 1978 and 2017; in fact, it has slightly decreased.
376 Chapter 9 Telling the Truth with Data Visualization

FIGURE 9.16 Adjusting Nominal Gasoline Prices for Inflation in Excel

Note that rows 11–39 are

hidden.

FIGURE 9.17 Adjusting for Inflation Shows that the Price per Gallon
of Gasoline in the United States Has Actually Decreased
between 1978 and 2017

Comparing Nominal and Inflation—Adjusted Gas Prices between 1978

and 2017
Price (per gallon)
4.00

3.50

3.00
Nominal Gas Price ($)
2.50

2.00

1.50

1.00

0.50 Inflation-Adusted Price (2017 $)

0.00
1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Year
Source: https://www.usinflationcalculator.com/gasoline-prices-adjusted-for-inflation/

We can adjust nominal values to account for inflation using any year in the time series.
Selection of the base year to use generally depends on the comparisons that are to be made.
 9-4 Deceptive Design 377

9-4 Deceptive Design

Effective data visualizations convey insights from data to the audience clearly and accu-
rately. We want to avoid deceptive designs that do not accurately communicate insights
from the data. It is easy to create deceptive data visualizations—even when the analyst is
not actively trying to deceive the audience. The previous sections in this chapter emphasize
the importance in ensuring the data are as accurate as possible. In this section we will show
how decisions related to the design of the chart can also greatly affect the insights con-
veyed to the audience and how to avoid deceptive design choices for data visualizations.

Design of Chart Axes

Choices related to the axes used in many different types of charts can greatly affect the
insights conveyed to the audience. Changing the lower and upper limits of either the verti-
cal or horizontal axes can completely change what the audience perceives from the chart.
Even changes to the vertical and horizontal size of some charts can change the insights
communicated to the audience. Therefore, the chart designer must be extremely careful
when making decisions related to the chart axes to avoid deceptive design practices.
It is common for political polling firms to gather data related to the way likely voters plan to
vote on proposed ballot issues. Consider the case in Franklin County, Maryland, where voters in
the next election will be asked to vote on a library levy. If approved, the levy will provide funds
from a property tax increase to rehabilitate two aging libraries in Franklin County. A local polit-
ical polling firm conducts a poll of likely voters in Franklin County to determine whether voters
are “For” or “Against” the levy. The results of the poll are shown in Figure 9.18 as a column
chart. This column chart appears to convey that many more voters are “Against” the levy than
“For” the levy. However, it is important to note that the vertical axis ranges only between 48%
and 52%. In fact, Figure 9.18 shows that 51% of likely voters are against the levy while 49%
are for the levy. Because the vertical axis is truncated to show only the range of 48–52%, this
difference appears to be much larger in Figure 9.18 than it actually is. Notice also that the title
used here, “Majority of Voters Against Library Levy,” emphasizes one particular interpretation
of the chart that may not be fully supported by the data.

FIGURE 9.18 Column Chart that Exaggerates the Difference in Proportion of Likely Voters
Against the Library Levy and the Proportion of Likely Voters For the Library Levy

Majority of Voters Against Library Levy

Percent of Likely Voters
52

51

50

49

48
For Against
378 Chapter 9 Telling the Truth with Data Visualization

Figure 9.19 shows a revised column chart for these data. There are several differences
between Figure 9.19 and Figure 9.18. The vertical axis now has a range of 0–60%, we have
added data labels on each column showing their value, and because these polling results
represent an estimated proportion from a sample, we have added errors bars to the figure.
Figure 9.19 more accurately conveys that there is little difference between the propor-
Confidence intervals are
tion of voters who are against the levy and the proportion of voters who are for the levy.
discussed in more detail in Figure 9.19 also shows that the 95% confidence intervals on the two proportions overlap,
Chapter 5. so we cannot claim that these results are statistically different at a 95% level of confidence.

FIGURE 9.19 Redesigned Column Chart for the Results from the Poll
Taken of Likely Voters that Shows There Is Little Dif-
ference between the Voters For the Levy and Voters
Against the Levy

Fate of Library Levy Uncertain

95% Confidence Interval
Percent of Likely Voters
60

49% 51%
50

40

30

20

10

0
For Against
Source: https://datahub.io/core/global-temp#data

Generally, column (and bar) charts should begin with zero as the minimum value on
the vertical (and horizontal) axis. Exceptions to this recommendation include cases where
the measure being displayed has some minimum value other than zero such as cash flows
that can take on negative values or scores on a standardized test where 400 is the minimum
score.
The title used in Figure 9.19, “Fate of Library Levy Uncertain,” also presents a different
insight than what is shown in Figure 9.18. Figures 9.18 and 9.19 illustrate the importance
of the chart title in determining which insights from the data are communicated to the
audience.
The minimum and maximum values used on the vertical axis of line charts can also
greatly influence the insights conveyed to the audience. Consider a researcher who is
examining changes to the global surface temperature of the earth over time. Figure 9.20
displays the average annual surface temperature for the earth from 1880 to 2016 in degrees
Celsius (°C).
The insight communicated about these data to the audience appears to be that average
annual global surface temperatures have basically remained unchanged between 1880 and
2016. However, Figure 9.21 displays the same data as Figure 9.20 but the range of the vertical
axis has been changed to have a minimum value of 13.5°C and maximum value of 14.5°C.
9-4 Deceptive Design 379

FIGURE 9.20 Line Chart Showing the Average Annual Global Surface
Temperature of the Earth between 1880 and 2016

Average Annual Global Surface Temperature

Temperature (°C)
50.0
45.0
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
1860 1880 1900 1920 1940 1960 1980 2000 2020
Year
Source: Data from https://datahub.io/core/global-temp#data and inspired by similar example on page 65 of Cairo,
How Charts Lie: Getting Smarter About Visual Information

FIGURE 9.21 Revised Line Chart Showing the Average Annual Global Sur-
face Temperature of the Earth between 1880 and 2016 with
Reduced Range of Values for the Vertical Axis

Average Annual Global Surface Temperature

Temperature (°C)
14.5
14.4
14.3
14.2
14.1
14.0
13.9
13.8
13.7
13.6
13.5
1860 1880 1900 1920 1940 1960 1980 2000 2020
Year
Source: https://datahub.io/core/global-temp#data

From Figure 9.21, we can see that there is actually a substantial amount of variabil-
ity present in these data compared to what was apparent in Figure 9.20. Figure 9.21 also
shows what appears to be an upward trend in the data after around 1920 that is not apparent
in Figure 9.20.
380 Chapter 9 Telling the Truth with Data Visualization

Preattentive attributes are For charts that use preattentive attributes of length and width such as bar and column
discussed in detail in Chapter 3.
charts, it is generally recommended that the vertical axis start at 0 unless there is a different
relative minimum value for the variable being used on the vertical axis. However, for charts
that use preattentive attributes related to orientation or spatial positioning such as line
charts and scatter charts, it is not always recommended to start the vertical axis at zero or
the relative minimum value. To communicate the information in the data through orienta-
tion or spatial positioning, it may be necessary to use a smaller range on the vertical axis to
illustrate the variability, trend, or correlation in the data.
Figure 9.22 shows two additional possible line charts to display the average annual global
surface temperatures over time. The charts in Figure 9.22 have identical ranges for the axes,
but they have different aspect ratios. The aspect ratio refers to the ratio of the width of a chart
to the height of a chart. In Figure 9.22a, the line chart is tall, but not very wide. Figure 9.22b
shows the same data in a line chart that is short but wide. In other words, Figure 9.22a has a
smaller aspect ratio than Figure 9.22b. Comparing these two different line charts shows that
using a line chart that is tall and narrow (smaller aspect ratio) can exaggerate trends in the
data, while using a line chart that is short and wide (larger aspect ratio) can disguise trends in
the data even when we have the same minimum and maximum vertical axis values.
The aspect ratio is discussed Figure 9.22 demonstrates the importance of considering the aspect ratio used in creating
in Chapter 6. data visualizations when showing them to an audience, as well as the importance of the
audience to consider the effect of the aspect ratio on the data visualizations when viewing
them. This can be particularly important, and challenging, when designing multiple data
visualizations that will be dynamically updated such as for data dashboards.

FIGURE 9.22 Two Different Line Charts for the Average Annual Global Surface Temperature
Data Showing the Effect of the Size of the Line Chart on the Insights Conveyed
to the Audience

Average Annual Global

Surface Temperature
Temperature (°C)
14.5
14.4
14.3
14.2
14.1
14.0
13.9
13.8
13.7
13.6
13.5
1870 1920 1970 2020
Year
(a)

Average Annual Global Surface Temperature

Temperature (°C)
14.5
14.3
14.1
13.9
13.7
13.5
1860 1880 1900 1920 1940 1960 1980 2000 2020
Year
(b)
Source: https://datahub.io/core/global-temp#data
9-4 Deceptive Design 381

Dual-Axis Charts
Sometimes it is necessary to display two different variables from a data set to the audience
where each variable has different units and/or different magnitudes. A common approach
to creating a chart for such a situation is to use what is known as a dual-axis chart. A
dual-axis chart makes use of a secondary axis to represent one of the variables so that both
variables can be shown on the same chart. However, in most cases, dual-axis charts are dif-
ficult for the audience to interpret, and there is often a better way to present the data.
Consider a case for which the audience is interested in comparing the gross-domestic prod-
uct (GDP) and unemployment rate of the United States since the year 2000. The GDP of the
United States is measured in the trillions of dollars while the unemployment rate is specified as
a percentage value of the eligible workforce. Figure 9.23 shows the data for both variables on
the same chart using the same vertical axis. Clearly this chart is not particularly informative for
the audience. The unemployment rate appears to be a flat line at zero because the vertical axis
must extend much higher even when the unit is expressed in billions of U.S. dollars. Further,
with only a single axis, it can be confusing for the audience to understand the units of measure
for the vertical axis because GDP and unemployment rate are measured in different units.

FIGURE 9.23 Displaying the Gross Domestic Product and Unemploy-

ment Rate in the United States since 2000 Using a Single
Vertical Axis

Comparing Gross Domestic Product (GDP) and Unemployment Rate in the

United States Since 2000
25000
GDP ($ billions)
20000

15000

10000

5000

Unemployment Rate %
0
2000 2005 2010 2015 2020
Year
Source: Bureau Labor Statistics and World Bank; https://data.bls.gov/timeseries/LNU04023554&series_id=LNU04000000&
series_id=LNU03023554&series_id=LNU03000000&years_option=all_years&periods_option=specific_periods&periods
=Annual+Data and https://data.worldbank.org/indicator/NY.GDP.MKTP.CD?end=2019&start=1960&view=chart

A common alternative to the chart shown in Figure 9.23 is to use a dual-axis chart. The
dual-axis chart makes use of a secondary vertical axis to represent one of the variables. In
this case, the secondary axis will represent the unemployment rate. The resulting dual-axis
chart is shown in Figure 9.24. Because we can now use two different vertical axes, the
unemployment rate no longer appears as a flat line. We can now see changes in both the
U.S. GDP and the unemployment rate on the chart and compare the two lines.
However, dual-axis charts can still be misleading and difficult for the audience to inter-
pret. Because the lines in Figure 9.24 appear together on the same chart, it is natural for the
audience to compare them to each other. At first glance, it can appear that the unemploy-
ment rate is higher than the GDP in certain years. The audience must correctly match each
line to the corresponding vertical axis, which requires considerable cognitive load.
382 Chapter 9 Telling the Truth with Data Visualization

It is also incorrect to interpret meaning from a comparison of the slopes of the lines
shown in Figure 9.24. Because each line is scaled to a different vertical axis, the audience
cannot directly compare the slopes of the lines. What appears to be a rapid increase in
unemployment rates cannot be directly compared to the corresponding increase or decrease
in GDP because the scales of the vertical axes are different. As shown previously in this
chapter, the slopes of the lines are dependent on the ranges chosen for the vertical axes. In
general, the magnitudes of the representation of the data for the different variables cannot
be compared directly because the audience must adjust each representation of the data by
the scale of the appropriate vertical axis.
Another possible point of confusion with dual-axis charts is that the audience is typi-
cally drawn to places where lines intersect. The natural inclination is to assume that the two
variables are equal at these points of intersection in Figure 9.24. However, this is incorrect
in a dual-axis chart since each variable is represented on a different vertical axis. In reality,
the lines do not intersect at all if they are shown on the same vertical axis as in Figure 9.23.

FIGURE 9.24 Using a Dual Axis Chart to Display the Gross Domestic
Product and Unemployment Rate in the United States
since 2000

Comparing Gross Domestic Product (GDP) and Unemployment Rate in the

United States Since 2000
GDP ($ billions) Percent Unemployed
25000 12
GDP
10
20000

8
15000
6
10000
4
Unemployment Rate
5000
2

0 0
2000 2005 2010 2015 2020
Year
Source: Bureau Labor Statistics and World Bank; https://data.bls.gov/timeseries/LNU04023554&series_id=LNU04000000&
series_id=LNU03023554&series_id=LNU03000000&years_option=all_years&periods_option=specific_periods&periods
=Annual+Data and https://data.worldbank.org/indicator/NY.GDP.MKTP.CD?end=2019&start=1960&view=chart

A simple alternative to using a dual-axis chart is to replace the dual-axis chart with two
charts where each variable is shown on a different chart. An example of this is shown in
Figure 9.25. Using two different charts to display the data helps the audience understand
that each variable uses a different vertical axis and that the magnitudes and slopes cannot
be compared directly. Because the horizontal axis is the same for each of these charts, it
can be helpful to stack the charts vertically so that the horizontal axes align. The downside
of using two charts is that it takes more space to display both charts.

Data Selection and Temporal Frequency

Choosing which data to include in a chart greatly affects the insights conveyed to the audi-
ence. This is particularly true for time series data collected over successive points in time.
Time series data are discussed Stock prices in finance, inventory levels at a retail store, water flow rates in a river, patient
in Chapter 6. occupancy levels in a hospital, political polling data collected over time, and many other
applications often rely on the use of time series data.
9-4 Deceptive Design 383

FIGURE 9.25 Using Two Different Charts to Display the Gross Domestic
Product and Unemployment Rate in the United States
since 2000

Gross Domestic Product (GDP) in the United States Since 2000

GDP ($ billions)
25000

20000

15000

10000

5000

0
2000 2005 2010 2015 2020
Year

Unemployment Rate in the United States Since 2000

Percent Unemployed
12

10

0
2000 2005 2010 2015 2020
Year

Source: Bureau Labor Statistics and World Bank; https://data.bls.gov/timeseries/LNU04023554&series_id=LNU04000000&

series_id=LNU03023554&series_id=LNU03000000&years_option=all_years&periods_option=specific_periods&periods
=Annual+Data and https://data.worldbank.org/indicator/NY.GDP.MKTP.CD?end=2019&start=1960&view=chart

Consider the charts shown in Figure 9.26, which appear to represent the share prices of
two stocks (A and B) for potential investment. Trendlines have also been added to these
charts to show the general linear trend in the data. The stock prices shown in Figure 9.26a
appear to be highly variable with a trendline that is basically flat. The stock prices shown
in Figure 9.26b appear to be more stable and have an increasing trend, and so may be per-
ceived as a superior potential investment. However, both charts actually represent the same
stock; both charts in Figure 9.26 have the same ending date, but the start dates are different.
Figure 9.26a shows the data for the last 90 days, and Figure 9.26b shows the data for the last
20 years. The audience would likely feel differently about the possibility of investing in a
stock whose price behaved as shown in Figure 9.26a than in a stock whose price has behaved
as shown in Figure 9.26b. This illustrates how changing the range of the dates on the horizon-
tal axis can greatly affect the insights conveyed to the audience. Figure 9.26 also reinforces
the importance of labeling vertical and horizontal axes. Without the horizontal axis labels in
Figure 9.26, it can be easy for an audience to misinterpret the insights from these charts.
384 Chapter 9 Telling the Truth with Data Visualization

FIGURE 9.26 Two Charts Showing Stock Share Prices and Associated
Trendlines over Different Time Frames

Potential Stock Investment A: Performance over Last 90 Days

Share Price ($)
36

35

34

33

32
(a)

Potential Stock Investment B: Performance over Last 20 Years

Share Price ($)
40

35

30

25

20

15

10

0
(b)
Source: Yahoo Finance, https://finance.yahoo.com/quote/%5EGSPC?p=^GSPC&.tsrc=fin-srch

Whether Figure 9.26a or Figure 9.26b is more appropriate depends on the needs of the
audience. If the audience does not intend to hold the stock for very long and is only inter-
ested in short-term performance of the share price, Figure 9.26a may be more relevant.
However, if the audience intends to invest for the long term, Figure 9.26b may be more
appropriate.
Next compare the share prices of the stocks shown in the charts in Figure 9.27. In these
charts, the same start and end dates are used for the horizontal axis in each chart. The stock
Volatility is a common displayed in Figure 9.27a appears to have much higher price volatility than the stock displayed
measure of the amount of in Figure 9.27b. Surprisingly, however, these charts both show share prices for the same stock.
change in a stock price.
While the horizontal axes have the same start and end dates in each chart, what varies here
is the temporal frequency with which the share price is plotted. The temporal frequency in
 9-4 Deceptive Design 385

a chart refers to the rate at which time-series data are displayed in a chart. In Figure 9.27a,
the share price is plotted daily. In Figure 9.27b, the share price is plotted every other month.
Therefore, the temporal frequency used in Figure 9.27a is 365 per year while the temporal
frequency in Figure 9.27b is 6 per year. Using every-other-month data in Figure 9.27b has
the effect of reducing the variability that is shown in the chart. In other words, Figure 9.27b
smooths out the data. This smoothing effect can also hide seasonal or cyclical effects that are
Temporal frequency is present in the data. Viewing time series data at a different temporal frequency can reveal pat-
discussed in Chapter 6.
terns that exist at that frequency but hide patterns that exist at other frequencies.

FIGURE 9.27 Comparing Stock Prices with the Same Start and End
Dates but Different Temporal Frequencies

Stock Values Shown with Temporal Frequency = 365 per Year

Share Price ($)
38
36
34
32
30
28
26
24
22
20

20
19

19
18

19

19

20

20

20
20

20
20

20

20

20

20

6/
3/

1/
7/

4/

5/

0/

8/

/2
/2

/0
/2

/0

/1

/1

/1

09
08

12
10

02

05

03

06
(a)

Stock Values Shown with Temporal Frequency = 6 per Year

Share Price ($)
38
36
34
32
30
28
26
24
22
20
19
19

20
20
19
19

20
18

20
20

20
20
20
20

20
20

1/
3/

6/
8/
5/
4/

0/
7/

/0
/2

/2
/1
/1
/0

/1
/2

12
08

09
06
05
02

03
10

(b)

Source: Yahoo Finance, https://finance.yahoo.com/quote/%5EGSPC?p=^GSPC&.tsrc=fin-srch

Figure 9.27 shows that even when time series charts have the same start and end dates
on the horizontal axes, the insights conveyed in the chart can still differ based on the tem-
poral frequency with which the data are shown. Charts that use a lower temporal frequency
will tend to reduce the apparent variability shown in the chart.
386 Chapter 9 Telling the Truth with Data Visualization

A general recommendation for time series is to collect the data at the highest possi-
ble temporal frequency. That is, data should be collected at as frequent time intervals as
possible. This is because it is easy to aggregate data, but it is usually difficult, or even
impossible, to disaggregate it. If we have data that are measured at the daily level, it is easy
to aggregate these data to weekly, monthly, quarterly, or annual levels. However, if the
data only exist at the annual level, it may be impossible to disaggregate to the quarterly,
monthly, weekly, or daily level.

Issues Related to Geographic Maps

Choropleth maps and The use of geographic maps for data visualization introduces additional possibilities
other geographic maps are for misleading the audience. Choropleth maps that use different shades of a color to
discussed in Chapters 2 and 6. represent quantitative variables are a common type of data visualization for exploring
and examining data related to different geographic regions. Choropleth maps are com-
monly used to examine the differences in many different economic and health variables
such as employment rates, income levels, cancer rates, political support, and average
life spans across geographic regions such as counties, states, regions, and countries.
Consider the choropleth map shown in Figure 9.28 in which color shading is used to
denote the estimated number of individuals living below the poverty line in each state
in the United States. This map appears to show that states such as California (CA),
Texas (TX), Florida (FL), and New York (NY) have the most poverty of any states.
While it is true in an absolute sense that these states have the highest number of people
living below the poverty line, it is also true that these states also have the highest pop-
ulations. Compare Figures 9.28 to 9.29, which shows a choropleth map that represents
the population in each state. Figure 9.28 and Figure 9.29 are extremely similar. In fact,
many choropleth maps that show the absolute number of people in almost any partic-
ular category will appear similar to Figure 9.29. Figure 9.28 does not provide much
insight into which states have the greatest challenges with poverty, because it is really
only showing which states have the largest populations.

FIGURE 9.28 Choropleth Map Showing Number of People Living in Poverty by State

Estimated Number of People Living in Poverty by State

Number in Poverty
WA 4,972,955
MT ME
ND
OR MN VT
ID NH
SD WI NY
MI 60,097
WY
IA PA
NE
NV OH
UT IL IN
CA CO WV
VA
KS MO KY
NC
TN
AZ OK
NM AR SC
MS AL GA
TX LA
AK FL
 9-4 Deceptive Design 387

FIGURE 9.29 Choropleth Map Showing Population by State

Estimated United States Population by State

2020 Population
WA 39,937,489
MT ME
ND
OR MN VT
ID
SD WI NY
MI 567,025
WY
IA PA
NE
NV OH
UT IL IN
CA CO WV
KS MO VA
KY
NC
TN
AZ OK
NM AR SC

MS AL GA
TX LA
AK FL

Figure 9.30 is a better choropleth map for communicating which states have the highest
number of people living in poverty relative to the overall state population. Figure 9.30 uses
the poverty rate rather than the total number of people living in poverty. The poverty rate
for each state is defined as shown in the equation below.
Number of people in poverty in a state
Poverty rate in a state 5 3 100%
Total population in a state

FIGURE 9.30 Choropleth Map Showing Poverty Rate by State

Estimated Poverty Rate by State

Poverty Rate
WA 19.1%
MT ND ME
OR MN VT
ID
SD WI NY
WY MI 7.3%
IA PA
NV NE
OH
UT IL IN
CA CO WV
KS MO VA
KY
NC
TN
AZ NM OK
AR SC
AL GA
MS
TX LA
AK
FL
388 Chapter 9 Telling the Truth with Data Visualization

In other words, the poverty rate measures the percentage of each state’s population that
is living in poverty. For Figure 9.30, we see that states such as Mississippi (MS), New
Mexico (NM), and Louisiana (LA) have the highest rates of poverty while California (CA),
Texas (TX), Florida (FL), and New York (NY) have lower poverty rates. In most cases, it is
best to use a value relative to the population of the region when creating choropleth maps.

N otes 1 C omments

1. The book How Charts Lie: Getting Smarter about Visual In general, we recommend that charts used to show cor-
Information by Alberto Cairo contains many additional relation between variables such as a scatter chart use an
examples of deceptive chart designs and how these decep- aspect ratio of 1:1 where the width is equal to the height.
tions can be remedied. For most other charts, it is recommended that the width
2. Dual-axis charts can be created in Excel by clicking the Insert is larger than the height, but there are exceptions to this
tab on the Ribbon, clicking the Insert Combo Chart button such as for slope charts, introduced in Chapter 7, where the
Create Custom Combo Chart...
in the Charts group and then selecting . width is usually smaller than the height.
To create a dual-axis chart, select a Chart Type for each vari- 4. Data labels can be added to a map in Excel by clicking on
able to be shown in the chart, and then select the check box the map, clicking on the Chart Elements button , click-
under Secondary Axis for one of the variables. The dual-axis ing the arrow next to Data Labels and selecting More
chart can be a single type of chart, such as the line chart shown Data Label Options… to open the Format Data Labels
in the example in this chapter, or it can combine chart types task pane. Then select the appropriate option under Label
such as a column chart and a line chart. Options to add data labels to the map.
3. The proper aspect ratio for a chart depends on the
exact chart and how it will be displayed to the audience.

S u mma r y

In this chapter, we have covered several common issues in data visualizations that can lead
to confusing and misleading interpretations. The goal of effective data visualization is to
convey insights accurately and truthfully in a way that requires as little cognitive load from
the audience as possible. Deceptive data visualizations are sometimes created intentionally
to mislead an audience, but more often it is caused by a lack of understanding of the needs
of the audience, poor data quality, or poor data visualization design choices.
Creating truthful data visualizations begins with having complete and accurate data. We
began this chapter with a description of some simple methods for identifying missing data
and data errors. The exact nature of missing data and data errors is specific to the type of
data and problem setting being analyzed. However, the methods introduced here are simple
enough to complete in Excel and can help mitigate some obvious cases of missing data and
data errors.
We discussed the importance of adjusting for inflation when dealing with time series
data that have been collected over long time periods. Failing to adjust for inflation can eas-
ily lead to incorrect insights and misleading data visualizations. We discussed the impor-
tance of considering choices related to the design of axes in charts. Using different ranges
of values for the axes, and even simply using different sizes for the chart, can greatly affect
the insights conveyed to the audience for the same data. We explained why the use of dual-
axis charts is often confusing or misleading for the audience, and we suggested the use of
two separate charts that each show a single variable instead. We also discussed how the
selection and frequency of time series data can change the insights conveyed to the audi-
ence. Finally, we covered some issues related to the use of geographic maps and how to
prevent misleading the audience when using these types of charts.
There are many other data visualization design decisions that can lead to misleading
charts. Decisions as simple as the wording used in the title of the chart can sometimes
mislead the audience. There is a fine line between influencing the audience by highlighting
 Problems 389

specific insights from the data and misleading the audience using deceptive design.
However, the goal of an effective data visualization is to convey insights to the audience as
truthfully as possible. The material covered in this chapter, and throughout this textbook,
should help create more truthful and more effective data visualizations.

G L O S S A RY

Aspect ratio The proportion between the chart’s width and its height.
Base year Arbitrary year chosen to be the common year to measure economic values such
as costs and prices to adjust for inflation.
Biased data Sample data that are not representative of the population that is under study.
Dual-axis chart A type of data visualization that makes use of a secondary axis to
represent one of the variables so that both variables can be shown on the same chart.
Inflation The general increase in prices over time.
Nominal values Raw values that have not been adjusted for inflation or other important
factors.
Price index Measure of the relative change in the price of a standard set of products and
services over time.
Real values Values that have been adjusted for inflation.
Selection bias Bias that occurs when data are drawn from a sample that has not been
properly randomized to represent the intended population.
Simpson’s paradox Subsets of data show a specific trend that either disappears or reverses
when the data are aggregated.
Survivor bias Bias that occurs when a sample data set consists of a disproportionately
large number of observations corresponding to positive outcomes for a particular event.
Temporal frequency The rate at which time-series data are displayed in a chart.

P R O B L E M S

Conceptual
1. Identifying Missing Data Which of the following Excel methods can be used to help
identify missing data? LO 1
i. Conditional formatting to highlight missing data values
ii. Using the COUNTIF() function to find the number of missing data values
iii. Sorting the data to find missing data values
iv. All of the above
2. Data Errors. Which of the following is true regarding data errors? LO 1
i. A data error is always identified by a unique numerical value such as 9999999.
ii. Any data that are determined to be outliers should be considered data errors and
should be removed.
iii. Identifying outliers in a data set can be helpful in uncovering data errors.
iv. Data errors occur only when data are collected manually.
3. Types of Data Bias. Match each of the following type of data bias with the correct
description. LO 2, 3

Type of Data Bias Description

Selection bias Occurs when subsets of data show a specific trend, but that trend dis-
appears, or reverses, when the data are aggregated.
Simpson’s paradox Occurs when a data set is made up of a disproportionately high number of
observations corresponding to positive outcomes from a particular event.
Survivor bias Occurs when data are drawn from a sample that had not been prop-
erly randomized to represent the intended population.
390 Chapter 9 Telling the Truth with Data Visualization

4. Potential Bias in Weight-Loss Study. A clinical trial has been conducted to evalu-
ate the efficacy of a new drug to enable weight loss for obese patients. A pool of 249
obese individuals are chosen for the study. Study participants must track their weight
at home daily to compute their body-mass index (BMI) and have a clinical evaluation
once per week at a local hospital over six months to complete the clinical trial. At the
end of six months, it is found that 47% of those who received the new drug completed
the clinical trial. Those who completed the clinical trial are found to have reduced their
BMI by 3.2 kg/m2, on average, over six months. Explain how these results could be
affected by bias and how that could affect the data. LO 2
5. Simpson’s Paradox in Baseball. In the sport of baseball, batting average is calculated
by dividing the number of hits a player achieves by the number of official at bats. The
following table shows the batting performance for Mike Legg and Edison Vasquez over
two consecutive seasons. LO 3

Season 1 Season 2
Player Hits At Bats Hits At Bats
Mike Legg 14 56 192 589
Edison Vasquez 112 418 57 149

a. Calculate the batting average for Mike Legg and Edison Vasquez in Season 1.
Which player has a higher batting average in Season 1?
b. Calculate the batting average for Mike Legg and Edison Vasquez in Season 2.
Which player has a higher batting average in Season 2?
c. Calculate the batting average for Mike Legg and Edison Vasquez over the combined
Seasons 1 and 2. Which player has the higher batting average over Seasons 1 and 2?
d. Explain how the results in parts a through c illustrate Simpson’s paradox.
6. Average Nominal Hourly Earnings. An economist is examining wage growth
in the United States. She has collected data on average nominal hourly earnings
for workers in the United States between 2006 and 2020, which are shown in the
following chart. Her conclusion is that hourly earnings for workers in the United
States have been steadily growing since 2006. How might this conclusion be incor-
rect, and how can she modify these data to investigate the appropriateness of this
conclusion? LO 4
Nominal Earnings in the United States between 2006 and 2020
Average Hourly Earnings ($)
35

30

25

20

15

10

0
13

20
10

11

12
09

18

19
08

14

16
15
06

07

17
20

20
20

20

20
20

20

20
20

20

20
20
20

20

20
1/

1/
1/

1/

1/
1/

1/

1/
1/

1/

1/
1/
1/

1/

1/
3/

3/
3/

3/

3/
3/

3/

3/
3/

3/

3/
3/
3/

3/

3/
 Problems 391

7. Meaning of Inflation. Which of the following are accurate statements regarding infla-
tion? (Select all that apply.) LO 4
i. Inflation refers to the tendency of prices to increase over time.
ii. Inflation only occurs in economies that are experiencing destabilizing events like
war or famine.
iii. Prices that are not adjusted for inflation are referred to as nominal prices.
iv. Failing to adjust for inflation in time series data that relates to monetary units can
lead to misleading data visualizations.
v. When creating a line chart for cost data over time, it does not matter if the data
are adjusted for inflation or not.
8. Political Polling Results in County Commissioner Election. Two candidates are
running for county commissioner in Bell County, Texas: Lisa Adamek and Rosemary
Andrews. The local newspaper has conducted a poll of likely voters to see where the
candidates stand in regards to the election. The local newspaper has analyzed the
results of its poll and posted the following chart to its website. LO 5

Poll for Bell County Commissioner Election

Percent of Likely Voters in Favor
54%

52%

50%

48%

46%
Lisa Adamek Rosemary Andrews

a. Why might this chart be misleading for the audience?

b. How could this chart be improved to be less misleading for the audience?
9. COVID-19 Positivity Rates in Ohio. An epidemiologist has collected data on
positivity rates for COVID-19 testing in the state of Ohio. The positivity rate is
defined as the number of tests that come back positive for the COVID-19 virus
divided by the total number of COVID-19 tests completed. The epidemiologist
has created the following chart showing positivity rates in Ohio between April and
October 2020. The audience for this chart is interested in trying to gain insights
into any overall increasing or decreasing trends in the positivity rate for COVID-19
during this time period. LO 5

Positivity Rate for COVID-19 in Ohio

Percent Positive Tests
100%

50%

0%
4/1/2020 5/1/2020 6/1/2020 7/1/2020 8/1/2020 9/1/2020 10/1/2020
Source: Data from https://covidtracking.com/data/download

How can this chart be improved to meet the needs of the audience?
392 Chapter 9 Telling the Truth with Data Visualization

10. Dual-Axis Charts. Which of the following statements are true of dual-axis charts?
(Select all that apply.) LO 6
i. Dual-axis charts use a secondary vertical axis to represent one of the variables
shown in the chart.
ii. Dual-axis charts require two different horizontal axes and two different vertical
axes for each variable displayed in the chart.
iii. Dual-axis charts can be used to display two different types of data visualizations
such as a line graph and a column graph on the same chart.
iv. Dual-axis charts can be difficult for the audience to interpret because the mag-
nitudes of the representation of the data cannot be compared directly without
adjusting for the scale of each vertical axis.
v. Dual-axis charts can only be created for pie charts and column charts.
11. Headcount and Revenue at Maximus Fashion. Maximus Fashion is a high-end
clothing store in Naperville, Illinois. The store manager would like to create a
data visualization that shows trends in the store’s headcount of retail associates
and revenues. For the last eight quarters, the manager has collected data on the
store’s headcount of retail associates (expressed in terms of full-time equivalent
[FTE] employees) and the store’s revenues. The manager has chosen to display
these data as the dual-axis chart that follows. The audience for this chart is
an informal advisory group for the store who has considerable experience in fash-
ion but little experience with financial analysis or data visualization in general.
LO 5, 6

Number of Retail Associates and Revenue at Maximus Fashion

Headcount (FTE) Revenue ($)
20 900,000
18 800,000
Revenue
16 700,000
14
600,000
Headcount
12
500,000
10
400,000
MaximusFashionChart 8
300,000
6
4 200,000

2 100,000

0 0
1 2 3 4 5 6 7 8
Quarter

a. What possible confusion could the use of this chart create for the audience?
b. How could the store manager redesign this chart to better present these data to the
audience?
12. Investing in Amazon.com Stock. Marisha Ray is a 30-year-old software engineer.
Marisha is considering investing in a large amount of Amazon.com stock that she
plans to hold until her retirement around age 60. Consider the following charts.
Each chart shows the performance of Amazon.com stock, but over different time
periods. One chart shows the performance of Amazon.com stock over the last five
days, and the other chart shows the performance of Amazon.com stock over the
last five years. Which chart would be the best chart to show to Marisha as she
considers investing in Amazon.com stock for her retirement savings? Why? LO 7
 Problems 393

Performance of Amazon.com Stock over Last 5 Days

Adjusted Closing Price ($)
Performance of Amazon.com Stock over Last 5 Days
3360
Adjusted Closing Price ($)
3340
3360

3320
3340

3300
3320

3280
3300

3260
3280

3240
3260

3220
3240

3200
3220

3180
3200

3160
3180
0 1 2 3 4 5
3160 Day
0 1 2 3 4 5
Source: https://finance.yahoo.com/quote/AMZN/history?period1=1445472000&period2=1603324800&
Performance of Amazon.com Stock over Last 5 Years
interval=1d&filter=history&frequency=1d&includeAdjustedClose=true Day
Adjusted Closing Price ($)
Performance of Amazon.com Stock over Last 5 Years
4000
Adjusted Closing Price ($)
4000
3500

3500
3000

3000
2500

2500
2000

2000
1500

1500
1000

1000
500

500
0
1 2 3 4 5
0 Year
1 2 3 4 5
Year
Source: https://finance.yahoo.com/quote/AMZN/history?period1=1445472000&period2=1603324800&
interval=1d&filter=history&frequency=1d&includeAdjustedClose=true

13. Communicating Insights with Time Series Charts. Which of the following properties
of a time series chart can influence the insights communicated to the audience? LO 7
i. The temporal frequency with which the data are shown
ii. The start and end date used on the horizontal axis for the chart
iii. The range of values displayed on the vertical axis
iv. All of the above
14. Poverty and Millionaires Choropleth Maps. The following choropleth maps show
the number of people living in poverty by state and the number of millionaires by state,
respectively. Interestingly, these choropleth maps appear to be similar; states with a
high level of millionaires also appear to be the same states with high levels of poverty.
One possible insight from comparing these two choropleth maps is that poverty and
high wealth appear to be positively correlated. LO 8
394 Chapter 9 Telling the Truth with Data Visualization

Estimated Number of People Living in Poverty by State

Number in Poverty
WA 4,972,955
MT ME
ND
OR MN VT
ID
SD WI NY
MI 60,097
WY
IA PA
NE
NV OH
UT IL IN
CA CO WV
KS MO VA
KY
NC
TN
AZ OK
NM AR SC
MS AL GA
TX LA
AK FL

Number of Millionaires by State

Number of Millionaires
WA 885,225
MT ME
ND
OR ID MN
SD WI NY
MI 14,989
WY
IA PA
NE
NV OH
UT IL IN
CA CO WV
KS VA
MO KY
NC
TN
AZ OK
NM AR SC
MS AL GA
TX LA
AK FL

a. Why might it be incorrect to conclude that higher rates of poverty occur in states
that have higher rates of wealth based on these two choropleth maps?
b. How would you improve these choropleth maps to provide a better comparison
between poverty rates and rates of high wealth by state?
Applications

15. Java Cup Taste Data. Huron Lakes Candies (HLC) has developed a new candy bar
called Java Cup that is a milk chocolate cup with a coffee-cream center. In order
to assess the market potential of Java Cup, HLC has developed a taste test and fol-
JavaCup
low-up survey. Respondents were asked to taste Java Cup and then rate Java Cup’s
taste, texture, creaminess of filling, sweetness, and depth of the chocolate flavor
of the cup on a 100-point scale. The taste test and survey were administered to 217
 Problems 395

randomly selected adult consumers. Data collected from each respondent are pro-
vided in the file JavaCup. LO 1
a. Are there any missing values in HLC’S survey data? If so, identify the respondents
for which data are missing and indicate which values are missing for each of these
respondents.
b. Are there any values in HLC’S survey data that appear to be erroneous? If so, iden-
tify the respondent for which data appear to be erroneous and indicate which values
appear to be erroneous for each of these respondents.
16. Major League Baseball Attendance. Marilyn Marshall, a professor of sports econom-
ics, has obtained a data set of home attendance for each of the 30 major league baseball
franchises for each season from 2010 through 2016. Dr. Marshall suspects the data,
AttendMLB
provided in the file AttendMLB, is in need of a thorough cleansing. You should also
find a reliable external source of Major League Baseball attendance for each franchise
between 2010 and 2016 to use to help you identify appropriate imputation values fro
data missing in the AttendMLB file. (Hint: ESPN.com contains attendance data for
Major League Baseball franchises.)
a. Are there any missing values in Dr. Marshall’s data? If so, identify the teams and sea-
sons for which data are missing and which values are missing for each of these teams
and seasons. use the reliable external source of Major League Baseball Attendance for
each franchise between 2010 and 2016 to find the correct value in each instance.
b. Are there any values in Dr. Marshall’s data that appear to be erroneous? If so, iden-
tify the teams and seasons for which data appear to be erroneous and indicate which
values appear to be erroneous for each of these teams and seasons. Use the reliable
external source of Major League Baseball attendance for each franchise between
2010 and 2016 to find the correct value in each instance.
17. Unemployment Rate and Delinquent Loans. The rate of unemployment is often cor-
related with the amount of delinquent loans because people have a more difficult time repay-
ing loans on time if they are unemployed. The file UnemploymentRate contains data on
UnemploymentRate
unemployment rates and percent of delinquent loans for 27 cities in the United States. LO 2
a. Create a scatter chart to examine the relationship between unemployment rate and
percent of delinquent loans in these cities. Does there appear to be a relationship
between unemployment rate and delinquent loans?
b. Based on the scatter chart you created in part a, are there any data points that would
cause concern as possibly being a data error?
18. Amount of Vacation Used at Consulting Firm. Guaraldi and Associates is a man-
agement consulting firm located in Manhattan. The firm is interested in examining the
relationship between the amount of vacation that their consultants use per year and
GuaraldiConsultants
how long the consultant has been with the firm. All consultants at Guaraldi and Asso-
ciates, other than the Managing Partners, are considered either Junior Consultants or
Senior Consultants based on their skills and expertise. The file GuaraldiConsultants
contains data on all Junior and Senior Consultants at Guaraldi and Associates including
the years of service the consultant has at the firm and the number of hours that the con-
sultant took as vacation last year. LO 4
a. Create a scatter chart and add a trendline to examine the relationship between years
of service and amount of vacation time used for all consultants at Guaraldi and
Associates. Based on the scatter chart, what appears to be the relationship between
years of service and amount of vacation time used?
b. Create a scatter chart for the same data, but this time differentiate the points in the
scatter chart based on whether the consultant is junior or senior. Do you see the
same relationship in this scatter chart individually for junior and senior consultants
as you saw in part a for all consultants?
c. Do the results in parts a and b provide an example of Simpson’s paradox? Why or
why not?
396 Chapter 9 Telling the Truth with Data Visualization

19. Average Nominal Hourly Earnings (Revisited). In this problem, we revisit the data
used in Problem 6. The file NominalWages contains the data that were used to create
the line chart shown in Problem 6. This file also contains price index data that can be
NominalWages
used to adjust the nominal hourly earnings in the file for inflation. Use the price index
data to adjust the nominal hourly earnings data for inflation using 3/1/2006 as the base
period. Create a line chart for the inflation-adjusted hourly earnings and compare it to
the line chart in Problem 6. What do the inflation-adjusted hourly earnings data suggest
about the growth of hourly wages between 2006 and 2020? LO 5
20. Political Polling Results in County Commissioner Election (Revisited). In this
problem, we revisit the chart from Problem 8. The chart shown in Problem 8 could
CountyCommissionerChart
be misleading for an audience comparing the polling results of the candidates Lisa
Adamek and Rosemary Andrews. Redesign this chart using the data in the file
CountyCommissionerChart to be a more effective data visualization for the polling
results for these two candidates. LO 6
21. COVID-19 Positivity Rates in Ohio (Revisited). In this problem, we revisit the chart
from Problem 9. The chart in Problem 9 could mislead an audience about trends related
to the positivity rate for COVID-19 tests in Ohio between April and October 2020. Use
PositivityOhioChart
the data in the file PositivityOhioChart to redesign the chart to present a more effec-
tive data visualization for an audience trying to gain insights into the positivity rate of
COVID-19 testing in Ohio. LO 6
22. Headcount and Revenue at Maximus Fashion (Revisited). In this problem, we
revisit the chart from Problem 11. The dual-axis chart used in Problem 11 is likely not
an effective chart for an audience with little experience in finance and data visualiza-
MaximusFashionChart
tion. Using the data in the file MaximusFashionChart, redesign this dual axis chart to
be more effective for the audience. LO 7
23. Poverty and Millionaires Choropleth Maps (Revisited). In this problem, we
revisit the charts from Problem 14. The choropleth maps in that problem show the
number of people living in poverty by state and the number of millionaires by state,
PovertyMillionaires
respectively. However, these charts could be misleading for an audience. The file
PovertyMillionaires contains data on the number of people living in poverty in each
state, the number of millionaires in each state, and the population of each state. LO 8
a. Create a choropleth map that shows the poverty rate as a percent of total state popu-
lation in each state. Which states have the highest poverty rates?
b. Create a choropleth map that shows the rate of millionaires as a percent of total
state population. Which states have the highest rate of millionaires?
c. Compare the choropleth maps in parts a and b with the maps in Problem 14. Do you
see the same relationship in the maps in parts a and b between the poverty rate and
rate of millionaires as is seen in the maps in Problem 14? Why or why not?
 1.1 Applications in Business and Economics 397

References
Alexander, M., and Walkenbach, J. Excel Dashboards and Jones, B. Avoiding Data Pitfalls. Wiley, 2020.
Reports. John Wiley & Sons, 2013. Gallo, C. The Storyteller’s Secret: From TED Speakers to
Benton, C. J. Excel Pivot Tables and Introduction to Business Legends, Why Some Ideas Catch On and Others
Dashboards: The Step-by-Step Guide Paperback. Amazon Don’t. St. Martin’s Publishing Group, 2016.
Digital Services, 2019. Goldmeier, J. Advanced Excel Essentials. Apress, 2014.
Berengueres, J., Fenwick, A., and Sandell, M. Introduction to Goldmeier, J., and Duggirala, P. Dashboards for Excel.
Data Visualization and Storytelling: A Guide for the Data Apress, 2015.
Scientist. Independently published, 2019. Knaflic, C. N. Storytelling with Data: A Data Visualization
Berinato, S. Good Charts: The HBR Guide to Making Smarter, Guide for Business Professionals. John Wiley and Sons,
More Persuasive Data Visualizations. Harvard Business 2019.
Review Press, 2016. Knaflic, C. N. Storytelling with Data: Let’s Practice! John
Berinato, S. Good Charts Workbook: Tips, Tools, and Exercises Wiley and Sons, 2019.
for Making Better Data Visualizations. Harvard Business Kriebel, A., and Murray, E. #MakeoverMonday: Improving
Review Press, 2019. How We Visualize and Analyze Data, One Chart at a Time.
Cairo, A. How Charts Lie. W. W. Norton and Company, 2019. John Wiley and Sons, 2018.
Cairo, A. The Truthful Art: Data, Charts, and Maps for Mollica, P. Color Theory: An Essential Guide to Color—From
Communication. Pearson Education, 2016. Basic Principles to Practical Applications. Walter Foster
Camm, J. D., Cochran, J. J., Fry, M. J., and Ohlmann, J. W. Publishing, 2013.
Business Analytics. 4th ed., Cengage, 2021. Page, S. E. The Model Thinker: What You Need to Know to
Camm, J. D., Fry, M. J., and Shaffer, J. A Practitioner’s Guide Make Data Work for You. Basic Books, 2018.
to Best Practices in Data Visualization. Interfaces, 47:6, Reynolds, G. Presentation Zen: Simple Ideas on Presentation
473–488, 2017. Design and Delivery. Pearson Education, 2009.
Choy, E. Let the Story Do the Work: The Art of Storytelling for Tufte, E. R. Envisioning Information. Graphics Press, 1998.
Business Success. AMACOM, 2017. Tufte, E. R. The Visual Display of Quantitative Information.
Edwards, B. Color: A Course in Mastering the Art of Mixing Graphics Press, 1983.
Colors. Penguin, 2004. Tufte, E. R. Visual Explanations. Graphics Press, 1997.
Evergreen, S. D. Effective Data Visualization: The Right Chart Urban, C. Advanced Excel for Productivity. Independently
for the Right Data. 2nd ed., SAGE Publications, 2019. published, 2016.
Few, S. Information Dashboard Design: Displaying Data for Wexler, S., Shaffer, J., and Cotgreave, A. The Big Book of
At-a-Glance Monitoring. Analytics Press, 2013. Dashboards: Visualizing Your Data Using Real-World
Few, S. Information Dashboard Design: The Effective Visual Business Scenarios. John Wiley and Sons, 2017.
Communication of Data. O’Reilly Media, 2006. Wilke, C. O. Fundamentals of Data Visualization: A Primer
Few, S. Now You See It: Simple Visualization Techniques for on Making Informative and Compelling Figures. O’Reilly,
Quantitative Analysis. Analytics Press, 2009. 2019.
 Index

A Chart type, selecting, 26–75

bar chart, 41
Color, traits frequently associated with
various colors, 133
Accounting bubble chart, 33–35 Colorblindness
data visualization in, 11–12 choropleth map, 42–44, 138–139, 269–272, definition of, 155
Afghanistan, GDP of, 77–78 338–339 neglecting, 153–156
Analogous colors, definition of, 42, 269 Color wheel, 131
definition of, 134 column chart, 39–41 definition of, 131
Analytical dashboards, definition of, 271 creating and editing charts in Excel, 29–32 Column chart, 4–5, 39–41
Analytics, 3–4 defining goal of data visualization, 28–29 definition of, 4, 39
definition of, 3 selecting an appropriate chart, 28–29 Combo chart
Arctic Death Spiral, 227 funnel chart, 52–53 avoiding, 57
Area chart, definition of, 57 line chart, 35–38 definition of, 57
Aristotle’s Rhetorical Triangle, 249–250 map, 42–47 Complementary colors, definition of, 134
Aspect ratio, definition of, 248, 332 geographic map, 42–44 Confidence interval, 201–205
Audience heat map, 44–45 definition of, 201
audience member analytical comfort levels, treemap, 45–47 Connection, 89–91
231–234 overview, 28, 59 definition of, 89
audience member needs, 229–231 scatter chart, 32–33 Contrast, insufficient, 151–153
know your audience, 229–234 stock chart, 51–52 Control chart, 16
summary guide to chart selection, 54–58 definition of, 16

B Excel’s recommended charts tool, 57–58

some charts to avoid, 55–57
Cool hues, definition of, 132
Correlation, definition of, 248
BAN (big associated number), 248 waterfall chart, 49–51 Crime dashboard, 269
definition of, 248 when to use tables, 47–49 Cross-sectional data, 9
Bar chart, 8 tables versus charts, 47–49 definition of, 9
definition of, 7, 41 Choropleth map, 42–44, 138–139, 269–272, Cross-tabulation, 242–247
Base year, definition of, 326 338–339 definition of, 242
Bell-shaped distribution, 197 definition of, 42, 269 Customization tools, definition of, 270
Benford’s Law, 11–12, 179–180 Clustered bar chart, definition of, 41
definition of, 179 Clustered column chart, 12, 40
definition of, 12, 40
D
Biased data, 321–326
definition of, 321 Cognitive load, definition of, 79, 135 Data, types of, 8–11
selection bias, 321–324 Color big data, 10–11
survivor bias, 324–326 definition of, 130 categorical data, 8
Big associated number (BAN), 248 excessive, 148–151 cross-sectional data, 9
definition of, 248 preattentive attribute, 81 quantitative data, 8
Big data, 10–11 Color, purposeful use of, 128–173 time series data, 9–10
definition of, 10 color and perception, 130–135 Data cleansing, definition of, 238
Bins, definition of, 177 color psychology, 132 Data dashboard, 264–300
Box and whisker chart, 197–199 color symbolism, 132 applications of, 267–269
definition of, 197 hue, 130–132 crime dashboards, 269
Box Office Mojo, 313 luminance, 130–132 donor dashboards, 269
Bubble chart, 33–35 perceived color, 132–135 human resource dashboards, 268
definition of, 33 saturation, 130–132 investment dashboards, 267
color schemes and types of data, 135–141 manufacturing dashboards, 267

C
categorical color schemes, 135–137 marketing dashboards, 268
diverging color schemes, 139–141 personal fitness/heath dashboards,
sequential color schemes, 137–139 268–269
Cartogram, 272–273
common mistakes in the use of color in school performance dashboards, 269
definition of, 272
data visualization, 146–156 technical support dashboards, 268
Categorical color scheme, definition
excessive color, 148–151 data dashboard taxonomies, 269–271
of, 135–136
inconsistency across related charts, 153 data updates, 270
Categorical data, 8
insufficient contrast, 151–153 organizational function, 270–271
definition of, 8
neglecting colorblindness, 153–156 user interaction, 270
Categorical variable, 234–236
not considering mode of delivery, 156 definition of, 5, 267
definition of, 176
unnecessary color, 146–147 design, 271–273
Chart axes, design of, 329–332
custom color using the HSL color common mistakes in, 290
Chart selection summary, 54–58
system, 141–146 considering needs of data dashboard’s
Excel’s recommended charts tool, 57–58
overview, 130, 156–157 users, 271
some charts to avoid, 55–57
400 Index

Data dashboard, (Continued) Data Visualization Makeover overview, 228–229, 255

data dashboard engineering, 271–273 age distribution of U.S. gymnasts, 175–176 storytelling and presentation design, 248–254
understanding purpose of data Arctic Sea ice volume, 227–228 Aristotle’s Rhetorical Triangle, 249–250
dashboard, 271 fast-food restaurant sales, 77–78 Freytag’s Pyramid, 250–253
overview, 266–267, 291 increasing youth participation in primary storyboarding, 253–254
principles of effective data dashboards, 267 voting, 227 storytelling with charts, 242–248
using Excel tools to build a data New York City comptroller, 27–28 choosing the correct chart to tell your
dashboard, 273–290 The Science of Hitting, 128–130 story, 242–245
final review of a data dashboard, 289–290 top-10 movies of all time, 313–315 using preattentive attributes to tell your
protecting a data dashboard, 289–290 Washington State Transportation story, 246–248
using PivotTables, PivotCharts, and Improvement Board (TIB), 265–266 Exploratory data analysis (EDA), definition
Slicers to build a data Deceptive design, 329–332 of, 228
dashboard, 274–288 data selection and temporal frequency, Exploring data visually, 226–283
Data-ink ratio, 91–98 334–338 analysis of missing data, 256–260
definition of, 92 design of chart axes, 329–332 exploring patterns associated with miss-
Data selection and temporal frequency, 334–338 dual-axis charts, 333–334 ing data, 258–260
Data visualization, introduction to, 2–25, 28–29 issues related to geographic maps, 338–340 types of missing data, 256–258
analytics, 3–4 Decluttering, 94–97 analyzing variables one at a time, 234–242
data visualization in practice, 11–18 definition of, 94 exploring a categorical variable, 234–236
accounting, 11–12 Dependent variable (y), definition of, 205 exploring a quantitative variable, 237–242
engineering, 16 Descriptive analytics, definition of, 4 organizing data to facilitate exploration,
finance, 12–13 Distributions, creating from data, 176–192 230–234
human resource management (HRM), frequency distributions for categorical overview, 228–234, 273–274
13–14 data, 176–179 relationships between variables, 242–256
marketing, 14 relative frequency and percent association between two quantitative
operations, 14–16 frequency, 179–180 variables, 247–256
sciences, 16–17 visualizing distributions of quantitative crosstabulation, 242–247
sports, 17 data, 181–192 visualizing geospatial data, 269–273
definition of, 3 Diverging color palette, definition of, 139 cartograms, 272–273
overview, 3, 18–19 Diverging color scheme, definition of, 139 choropleth maps, 269–272
selecting an appropriate chart, 28–29 Donor dashboards, 269 visualizing time series data, 260–269
types of data, 8–11 Dot matrix chart, definition of, 238 highlighting patterns in time series data,
big data, 10–11 Dow Jones Industrial Index (DJI), 8 262–265
categorical data, 8 Drilling down, definition of, 270 rearranging data for visualization, 266–269
cross-sectional data, 9 Dual-axis chart, 333–334 viewing data at different temporal fre-
quantitative data, 8 definition of, 333 quencies, 260–262
time series data, 9–10 Dynamic dashboard, definition of, 270 Eye travel, minimizing, 98–100
why visualize data?, 4–8
data visualization for explanation, 7–8
data visualization for exploration, 4–7
E F
Data visualization and design, 76–127 EDA (exploratory data analysis), definition Finance, data visualization in, 12–13
choosing a font for text, 100–101 of, 228 First-digit law, 11–12
common mistakes in data visualization Empathy, definition of, 236 Flicker, definition of, 87
design, 102–110 Enclosure, definition of, 89 Font choice for text, 100–101
too many attributes, 108 Engineering, data visualization in, 16 Form, 81–84
trying to display too much information, Equal-area cartogram, 272–273 definition of, 81
104–105 definition of, 272 Frequency distribution, 176–179
unnecessary use of 3D, 109–110 Ethos, definition of, 249 definition of, 176
using Excel default settings for charts, Excel tools Frequency polygon, 186–188
106–108 build data dashboards with, 273–290 definition of, 186
wrong type of visualization, 102–104 final review of a data dashboard, 289–290 Freytag’s Pyramid, 250–253
data-ink ratio, 91–98 reading and editing charts in, 29–32 definition of, 250
Gestalt principles, 88–91 using default settings for charts (common Funnel chart, 14–15, 52–53
connection, 89–91 mistake in data visualization and definition of, 14, 52
enclosure, 89 design), 106–108
proximity, 88–89
similarity, 88
Excessive color, 148–151
Explaining visually to influence with
G
minimizing eye travel, 98–100 data, 226–263 Geographic map, 42–44
overview, 78, 111 know your audience, 229–234 definition of, 42
preattentive attributes, 78–88 audience member analytical comfort issues related to, 338–340
color, 81 levels, 231–234 Geospatial data
form, 81–84 audience member needs, 229–231 definition of, 269
length and width, 84–87 know your message, 234–242 visualizing, 269–273
movement, 87–88 empathizing with data, 236–242 cartogram, 272–273
spatial positioning, 87 who helps the decision maker?, 235–236 choropleth map, 269–272
 Index 401

Gestalt principles, 88–91 Long-term memory, definition of, 79 Percent frequency, definition of, 179
connection, 89–91 Luminance, 130–132 Percentile, definition of, 196
definition, 88 definition of, 81, 131 Perception and color, 130–135
enclosure, 89 Lurking variable, definition of, 249 color psychology, 132
proximity, 88–89 color symbolism, 132
similarity, 88
M hue, 130–132
luminance, 130–132

H Manufacturing dashboards, 267

Maps, 42–47
perceived color, 132–135
saturation, 130–132
Heat map, 44–45 geographic map, 42–44 Personal fitness/heath dashboards, 268–269
definition of, 44 heat map, 44–45 Pie chart, 27
Hierarchical data, definition of, 45 treemap, 45–47 avoiding, 55
Hierarchical filtering, definition of, 270 Margin of error, definition of, 201 Pivot chart, definition of, 273
High-low-close stock chart, 13, 51–52 Marketing, data visualization in, 14 Pivot table, definition of, 273
definition of, 13, 51 Marketing dashboards, 268 PivotTables, PivotCharts, and Slicers to build
Histogram, 181–186 MAR (missing at random), definition of, a data dashboard, 274–285
definition of, 181 257 Polar Science Center (University of
Horizontal bar chart, 28 MCAR (missing completely at random), Washington), 227
HSL color system, 141–146 definition of, 257 Population, definition of, 176
Hue, 130–132 Mean, definition of, 193 Power law, definition of, 179
definition of, 130 Median, definition of, 193 Preattentive attributes, 78–88
primary, 130–131 Message color, 81
Human resource dashboards, 268 empathizing with data, 236–242 definition of, 79
Human resource management (HRM) know your message, 234–242 form, 81–84
data visualization in, 13–14 who helps the decision maker?, 235–236 length and width, 84–87
Missing at random (MAR), definition of, movement, 87–88
257 spatial positioning, 87
I Missing completely at random (MCAR), Prediction interval, definition of, 205
definition of, 257 Predictive analytics, definition of, 4, 205
Iconic memory, definition of, 79
Missing data Predictive models, uncertainty in, 205–211
Illegitimately missing data, definition of,
exploring patterns associated with, 258–260 illustrating prediction intervals for a simple
257
identifying data errors, 318–321 linear regression model, 205–208
Independent variable (x), definition of, 205
identifying missing data, 315–318 illustrating prediction intervals for a time
Inflation
types of, 256–258 series model, 208–211
adjusting for, 326–328
Missing not at random (MNAR), definition Prescriptive analytics, definition of, 4
definition of, 326
of, 257 Price index, definition of, 326
Interactive dashboard, definition of, 270
Mode, definition of, 193 Primary hues, definition of, 131
Interquartile range (IQR), 239
Mode of delivery, 156 Probability distribution, definition of, 179
definition of, 196
Moneyball (Lewis), 17 Proximity, 88–89
Investment dashboards, 267
Motion, definition of, 87 definition of, 88
Movement, 87–88
J Moving average, definition of, 263
Multivariate analysis, 242
Q
Jittering, 191–192 definition of, 242 Quantitative data, 8
definition of, 191 definition of, 8
N visualizing distributions of, 181–192
K Nominal values, definition of, 326
Quantitative scales, definition of, 248
Quantitative variables, 237–242
Kernal density chart Noninteractive dashboard, definition of, association between, 247–256
definition of, 192 270 definition of, 176
Key performance indicator (KPI) Quartile, definition of, 196
definition of, 267
O
R
L Occlusion, definition of, 191
Operational dashboards, definition of, 270 Radar chart, 57
Legitimately missing data, definition of, 256 Operations, data visualization in, 14–16 avoiding, 55–57
Length, 84–87 Ordinal variables, definition of, 236 definition of, 55
definition of, 84 Orientation, definition of, 82 Random variable, definition of, 176
Lewis, Michael, 17 Outlier, definition of, 198 Range, definition of, 194
Linear regression, 6 Relative frequency
Line chart, 35–38
definition of, 35
P definition of, 179
and percent frequency, 179–180
Location, measures of, 193–194 Pathos, definition of, 249–250 RGB color model and color wheel, 130–131
Logos, 249 Perceived color, 132–135 Rhetorical Triangle, 249–250
definition of, 249 Percent frequency, 179–180 definition of, 249
402 Index

S Stock chart, 51–52

definition of, 51 U
Sample, definition of, 176 Storyboard, definition of, 253 Univariate analysis, definition of, 234
Sample statistics, uncertainty in, 200–205 Storytelling Unnecessary color, 146–147
displaying a confidence interval on a mean, with charts, 242–248 Unstacked data, definition of, 266
201–203 choosing the correct chart to tell your
displaying a confidence interval on a
proportion, 203–205
story, 242–245
using preattentive attributes to tell your V
Sankey chart, 86–87 story, 246–248 Variability
definition of, 86 definition of, 228 definition of, 260
San-serif fonts presentation design and, 248–254 measures of, 194–197
definition of, 101 Aristotle’s Rhetorical Triangle, 249–250 visualizing, 174–225
Saturation, 130–132 Freytag’s Pyramid, 250–253 creating distributions from data, 176–192
definition of, 81, 131 storyboarding, 253–254 frequency distributions for categorical
Scatter chart, 7, 32–34, 87, 89–90 Strategic dashboards, definition of, 271 data, 176–179
definition of, 6, 32 Strip chart, 189–192 relative frequency and percent
Scatter-chart matrix definition of, 189 frequency, 179–180
definition of, 252, 254 Survivor bias, 324–326 visualizing distributions of
School performance dashboards, 269 definition of, 324 quantitative data, 181–192
Sciences, data visualization in, 16–17 overview, 176, 211
Seasonality, definition of, 260
Selection bias, 321–324
T statistical analysis of distributions of
quantitative variables, 193–200
definition of, 321 Table lens, definition of, 254 box and whisker charts, 197–199
Sequential color palette, definition of, 137 Tables, when to use, 47–49 measures of location, 193–194
Sequential color schemes, definition of, 137 charts versus tables, 47–49 measures of variability, 194–197
Serif, definition of, 101 Tactical dashboards, definition of, 271 uncertainty in predictive models, 205–211
Serif font, definition of, 101 Tall data, definition of, 229 illustrating prediction intervals for a
Shape, 83–84 Technical support dashboards, 268 simple linear regression model, 205–208
definition of, 83 Telling the truth, designing data visualizations illustrating prediction intervals for a
Short-term memory, definition of, 79 to, 312–348 time series model, 208–211
Shot chart, 17–18 adjusting for inflation, 326–328 uncertainty in sample statistics, 200–205
definition of, 17 biased data, 321–326 displaying a confidence interval on a
Similarity, 88 selection bias, 321–324 mean, 201–203
definition of, 88 survivor bias, 324–326 displaying a confidence interval on a
Simple linear regression, definition of, 205 deceptive design, 329–332 proportion, 203–205
Simpson’s paradox, 323 data selection and temporal frequency, Variables
definition of, 323–324 334–338 analyzing one at a time, 234–242
Size, 83–84 design of chart axes, 329–332 exploring a categorical variable, 234–236
definition of, 83 dual-axis charts, 333–334 exploring a quantitative variable, 237–242
Skewness, 184–186 issues related to geographic maps, relationships between, 242–256
definition of, 184 338–340 association between two quantitative
Slicer missing data and data errors, 315–321 variables, 247–256
definition of, 273 identifying data errors, 318–321 crosstabulation, 242–247
linking to multiple PivotTables, identifying missing data, 315–318 Variation, definition of, 176
285–288 overview, 315, 340–341 Violin chart, 200
Slope chart, definition of, 243 Temporal frequency, definition of, 261, definition of, 200
Spaghetti chart, 17 336–337 Visual perception, 78–79
definition of, 17 Three D (3D), unnecessary use of, 109–110 definition of, 78
Sparkline, definition of, 266–267 Time interval widget, definition of, 270
Spatial positioning, 87
definition of, 87
Time series chart, 208–209
definition of, 208
W
Sports, data visualization in, 17 Time series data, 9–10 Wall Street Journal, The, 7
Spurious relationship, definition of, 249 definition of, 9, 208, 260 Warm hues, definition of, 32
Stacked bar chart, definition of, 41 Time series data, visualizing, 260–269 Washington Post, The, 175, 227
Stacked column chart, 13–14, 40–41 highlighting patterns in time series data, Waterfall chart, 49–51
definition of, 13, 40 262–265 definition of, 49
Stacked data, definition of, 266 rearranging data for visualization, White space, 93
Standard deviation, 195–196 266–269 definition of, 93
definition of, 195 viewing data at different temporal Wide data, definition of, 229
Static dashboard, definition of, 270 frequencies, 260–262 Width, 84–87
Statistical analysis of distributions of Treemap, 45–47 definition of, 84
quantitative variables, 193–200 definition of, 45 Williams, Ted, 129–130
box and whisker charts, 197–199 Trellis display, 188–189, 263 Word cloud, definition of, 10
measures of location, 193–194 definition of, 189
measures of variability, 194–197
Statistical inference, 200
Trend, definition of, 260
Truth. See Telling the truth, designing data
Y
definition of, 200 visualizations to Yahoo! Finance, 13