Statistics for

Business and Economics

H
AN

H
C

AN
C

H
O

AN
G

C
N

C
G

C
N

O
G
N
H
AN
C

H
AN
C

Chapter 2

H
O

AN
C
G

NH
N

C
DESCRIPTIVE STATISTICS
CA

C
C

G
O

O
N
NG

G
N
 Chapter Goals

H
After completing this chapter, you should be able to:
AN
n Create and interpret graphs to describe categorical

H
C

AN
variables:
C

H
O

AN
frequency distribution, bar chart, pie chart, Pareto diagram
G

C
N

Create a line chart to describe time-series data

C
n

C
Create and interpret graphs to describe numerical

O
n

G
variables:

N
H

n frequency distribution, histogram, ogive, stem-and-leaf display

AN

n Construct and interpret graphs to describe relationships

C

between variables: H
AN
C

H
O

Scatter plot, cross table

AN
C
G

n
N

Describe appropriate and inappropriate ways to display

C

C
n
O

C
data graphically
G

O
N

G
N
 Chapter Goals
(continued)

H
After completing this chapter, you should be able to:
AN

H
C
Compute and interpret the mean, median, and mode for a

AN
n
C

H
O

set of data

AN
G

C
N

C
n Find the range, variance, standard deviation, and

C
N
coefficient of variation and know what these values mean

O
G
N
Apply the empirical rule to describe the variation of
H

n
AN

population values around the mean

C

H
AN
C

Explain the weighted mean and when to use it

H
n
O

AN
C
G

Explain how a least squares regression line estimates a

N

C
n
O

C
linear relationship between two variables
G

O
N

G
N
 Chapter Topics

H
n Measures of central tendency, variation, and
AN

H
shapeC

AN
C

H
Mean, median, mode, geometric mean
O

C
n

AN
G

C
N

Quartiles

C
n

C
N
Range, interquartile range, variance and standard

O
n

G
deviation, coefficient of variation

N
H

Symmetric and skewed distributions

AN

n
C

H
Population summary measures
AN
C

H
O

AN
C
G

n Mean, variance, and standard deviation

N

C
O

The empirical rule and Bienaymé-Chebyshev rule

C
n
G

O
N

G
N
 Chapter Topics

H
(continued)
Five number summary and box-and-whisker
AN
n

H
C
plots

AN
C

H
O

C
Covariance and coefficient of correlation

AN
G

C
N

C
G
Pitfalls in numerical descriptive measures and

C
n

O
G
ethical considerations

N
H
AN
C

H
AN
C

H
O

AN
C
G
N

C
O

C
G

O
N

G
N
 2.1 Summarising data for a categorical variable

H
AN

H
Categorical

AN

AN
C

C
Data
C
O

C
O
G

O
G
N

H
N

AN
C
Tabulating Data Graphing Data

C
H

O
AN

G
H

N
AN

C
Frequency
C

Distribution C
Bar Pie Pareto
O
C

G
Chart Chart Diagram
O

Table
N
G
N
 2.1.1 The Frequency Distribution Table

Summarize data by category

H
AN

H
AN

AN
C

Example: Hospital Patients by Unit

C
C
O

C
Hospital Unit
O Number of Patients
G

O
G
N

H
N

AN
Cardiac Care 1,052

C
C
Emergency 2,245

O
AN

G
H

Intensive Care 340

N
AN

C
Maternity 552
C

C
O
C

Surgery 4,630
G
O

N
G

(Variables are
N

categorical)
 2.1.2 Charts
Bar and Pie Charts
H
AN

H
AN

AN
Bar charts and Pie charts are often used
C

C
C

for qualitative (category) data

O

C
O
G

O
G
N

H
N

AN
Height of bar or size of pie slice shows the

C
n

C
frequency or percentage for each

O
AN

G
H

category

N
AN

C
C

C
O
C

G
O

N
G
N
 Bar Chart Example
Hospital Number

H
Unit AN of Patients

H
AN

AN
C

Cardiac Care 1,052

C
C

Emergency 2,245 Hospital Patients by Unit

O

C
O
G

5000

O
Intensive Care 340
G
N

H
N
Maternity 552

AN
4000

patients per year

Surgery 4,630
Number of

C
3000

C
H

O
2000

AN

G
H

N
AN

C
1000
C

C
O
C

0
G
O

Cardiac

Emergency

Intensive

Surgery
Maternity
N
G

Care

Care
N
 Pie Chart Example
Hospital Number % of

H
Unit AN of Patients Total

H
Hospital Patients by Unit

AN

AN
Cardiac Care 1,052 11.93
C

C
C

Emergency 2,245 25.46 Cardiac Care

O

C
12%
Intensive Care 340 3.86
O
G

O
G
N

H
Maternity 552 6.26
N

AN
Surgery 4,630 52.50

C
C
Emergency

O
Surgery 25%

AN

G
H

53%

N
AN

C
C

C Intensive Care
O
C

G
4%
O

(Percentages
N

Maternity
G

are rounded to
N

the nearest 6%
percent)
 Pareto Diagram
Used to portray categorical data
H
n AN

H
AN

AN
A bar chart, where categories are shown in
C

C
C
O

C
descending order of frequency
O
G

O
G
N

H
N

AN
n A cumulative polygon is often shown in the

C
same graph

C
H

O
AN

G
H

Used to separate the “vital few” from the “trivial

N
AN

C
C

many” C
O
C

G
O

N
G
N
 Pareto Diagram Example

H
AN

H
Example: 400 defective items are examined

AN

AN
C

for cause of defect:

C
C
O

C
O
G

O
Source of
G
N

H
N

AN
Manufacturing Error Number of defects

C
Bad Weld 34

C
H

O
Poor Alignment 223

AN

G
H

N
AN

Missing Part 25

C
C

C
Paint Flaw O 78
C

G
O

Electrical Short 19
N
G
N

Cracked case 21
Total 400
 Pareto Diagram Example(continued)
Step 1: Sort by defect cause, in descending order
H
Step 2: Determine % in each category
AN

H
AN

AN
C

C
C

Source of
O

C
O
G

O
Manufacturing Error Number of defects % of Total Defects
G
N

H
N

AN
Poor Alignment 223 55.75

C
Paint Flaw 78 19.50

C
H

O
Bad Weld 34 8.50

AN

G
H

N
AN

Missing Part 25 6.25

C
C

C
Cracked case 21O 5.25
C

G
O

Electrical Short 19 4.75

N
G
N

Total 400 100%

 Pareto Diagram Example
(continued)
H
AN

H
Step 3: Show results graphically

AN

AN
C

C
% of defects in each category O
C

C
Pareto Diagram: Cause of Manufacturing Defect
O
G

O
60% 100%
G
N

H
N

N
90%

AN
cumulative % (line graph)
50%
80%

C
C
70%
(bar graph)

40%

O
AN
60%

G
H

N
AN

30% 50%

C
C

C
40%
O
C

20%
G
30%
O

N
G

20%
N

10%

10%

0% 0%
Poor Alignment Paint Flaw Bad Weld Missing Part Cracked case Electrical Short
1.4 Graphs for Time-Series Data
n A line chart (time-series plot) is used to show
H
AN
the values of a variable over time

H
AN

AN
C

C
C
O

C
O
G

O
n Time is measured on the horizontal axis
G
N

H
N

AN
C
The variable of interest is measured on the

C
n

O
AN

G
H

vertical axis

N
AN

C
C

C
O
C

G
O

N
G
N
 N
G
O
C
C
AN
N Thousands of subscribers
H

50
100
150
200
250
300
350

0
G
1990
O N
C G
1991
C O
1992 C
AN
1993 H C
1994
AN
1995
H
1996

1997N
1998
G
O
1999 C
2000
C N
2001
AN G
H O
Magazine Subscriptions by Year

2002 C
2003 C
2004
AN
H
Line Chart Example

2005

2006
N
G
O
C
C
AN
H
2.2 Summarising data for a quantitative variable
using tables and graphics

H
AN

H
Numerical Data

AN

AN
C

C
C
O

C
O
G

O
G
N

H
N

AN
C
Frequency Distributions Stem-and-Leaf

C
and

H
Display

O
AN

G
H

Cumulative Distributions

N
AN

C
C

C
O
C

G
O

Histogram Ogive
N
G
N
 2.2.1 The Frequency Distribution Table

What is a Frequency Distribution?

H
AN

H
AN

AN
C

A frequency distribution is a list or a table …

C
n

C
C
O

C
O
G

O
containing class groupings (categories or
G
N

G
n

H
N

AN
ranges within which the data fall) ...

C
C
and the corresponding frequencies with which

O
n

AN

G
H

data fall within each class or category

N
AN

C
C

C
O
C

G
O

N
G
N
 Why Use Frequency Distributions?

A frequency distribution is a way to

H
n AN

H
AN

AN
summarize data
C

C
C
O

C
The distribution condenses the raw data
O
G

O
n
G
N

H
into a more useful form...
N

AN
C
and allows for a quick visual interpretation

C
n

O
AN

G
H

of the data

N
AN

C
C

C
O
C

G
O

N
G
N
 Class Intervals
and Class Boundaries
H
AN

H
AN

AN
Each class grouping has the same width
C

C
C
O

Determine the width of each interval by

C

C
n
O
G

O
G
N

H
largest number - smallest number
N

AN
w = interval width =

C
number of desired intervals

C
H

O
AN

G
H

n Use at least 5 but no more than 15-20 intervals

N
AN

C
C

n Intervals never overlap O

C
C

G
O

Round up the interval width to get desirable

N
G

n
N

interval endpoints
 Frequency Distribution Example

H
AN

H
AN
Example: A manufacturer of insulation randomly

AN
C

C
selects 20 winter days and records the daily
C
O

C
high temperature
O
G

O
G
N

H
N

AN
24, 35, 17, 21, 24, 37, 26, 46, 58, 30,

C
C
H

O
32, 13, 12, 38, 41, 43, 44, 27, 53, 27

AN

G
H

N
AN

C
C

C
O
C

G
O

N
G
N
 Frequency Distribution Example
(continued)

H
AN

H
AN

AN
n Sort raw data in ascending order:
C

C
C

12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58
O

C
O
G

O
Find range: 58 - 12 = 46
G
N

G
n

H
N

AN
n Select number of classes: 5 (usually between 5 and 15)

C
C
Compute interval width: 10

O
n (46/5 then round up)

AN

G
H

N
AN

Determine interval boundaries: 10 but less than 20, 20 but

C
n
C

C
O
less than 30, . . . , 60 but less than 70
C

G
O

Count observations & assign to classes

G

n
N
 Frequency Distribution Example
(continued)
Data in ordered array:
H
AN

H
AN

AN
12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58
C

C
C
O

C
O
G

O
Relative
Interval Frequency
G
N

Percentage

H
Frequency
N

AN
10 but less than 20 3 .15 15

C
C
20 but less than 30 6 .30 30

O
AN

G
H

30 but less than 40 5 .25 25

N
AN

C
C

40 but less than 50 4 C .20 20

O
C

G
O

50 but less than 60 2 .10 10

N
G
N

Total 20 1.00 100

 2.2.2 Charts
n Histogram
H
AN

H
n A graph of the data in a frequency distribution

AN

AN
C

is called a histogram

C
C
O

C
O
G

O
n The interval endpoints are shown on the
G
N

H
N

AN
horizontal axis

C
C
n the vertical axis is either frequency, relative

O
AN

G
H

frequency, or percentage

N
AN

C
C

C
Bars of the appropriate heights are used to
n
O
C

G
O

represent the number of observations within

N
G
N

each class
 Histogram Example

H
Interval AN Frequency

H
Histogram : Daily High Tem perature

AN

AN
C
10 but less than 20 3

C
7

C
C

20 but less than 30 6 6

O

C
30 but less than 40 5
O 6
G

O
40 but less than 50 4 5
G
N

H
5
N

N
50 but less than 60 2

AN
4
Frequency
4

C
3

C
3

O
2

AN

G
H

N
AN

C
1
C

0
C 0
O
C

0
G
(No gaps
O

N
G

between 0 0 10 10 2020 30 30 40 40 50 50 60 60 70
N

bars) Temperature in Degrees

 Questions for Grouping Data
into Intervals
H
AN

H
AN

AN
1. How wide should each interval be?
C

C
C

(How many classes should be used?)

O

C
O
G

O
G
N

H
2. How should the endpoints of the
N

AN
n
intervals be determined?

C
C
Often answered by trial and error, subject to

O
n

AN

G
H

user judgment

N
AN

C
The goal is to create a distribution that is
C

C
n
O
neither too "jagged" nor too "blocky”
C

G
O

N
G

n Goal is to appropriately show the pattern of

N

variation in the data

 How Many Class Intervals?

H
AN

H
AN

AN
Many (Narrow class intervals)
C

C
C

may yield a very jagged distribution

O

C
n

O
G

O
with gaps from empty classes
G
N

G
5

H
Freque
N

N
Can give a poor indication of how 0

AN
ncy
n

12

20

28

36

44

52
Temperature

6
frequency varies across classes

C
C
H

O
AN

G
H

Few (Wide class intervals)

N
AN

n 12

C
10
may compress variation too much and
C

C
n 8

Frequency
O
C

yield a blocky distribution 6

G
O

4
N
G

n can obscure important patterns of 2

N

variation. 0
0 30 60 More
Temperature
(X axis labels are upper class endpoints)
 The Cumulative Frequency Distribuiton
Data in ordered array:
H
AN

H
12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58

AN

AN
C

C
C
O

C
O Cumulative Cumulative
G

O
Class Frequency Percentage
G
N

G
Frequency Percentage

H
N

AN
10 but less than 20 3 15 3 15

C
C
20 but less than 30 6 30 9 45

O
AN

G
H

30 but less than 40 5 25 14 70

N
AN

C
40 but less than 50 4 20 18 90
C

C
O
C

50 but less than 60 2 10 20 100

G
O

N
G

Total 20 100
N
 2.2.3 The Ogive
Graphing Cumulative Frequencies
H
AN

H
Upper

AN

AN
interval Cumulative
C
Interval endpoint Percentage

C
C

Less than 10 10 0
O

C
O
G

O
10 but less than 20 20 15
G
N

G
Ogive: Daily High Temperature

H
20 but less than 30 30 45
N

AN
30 but less than 40 40 70

C
40 but less than 50 50 90 100
Cumulative Percentage

C
50 but less than 60 60 100

H
80

O
AN

G
H

60

N
AN

C
C

C
40 O
C

G
O

20
N
G
N

0
10 20 30 40 50 60
Interval endpoints
 2.2.4 Stem-and-Leaf Diagram

H
AN

H
AN

AN
C

A simple way to see distribution details in a

C
C

n
O

C
data set O
G

O
G
N

H
N

AN
C
METHOD: Separate the sorted data series

C
H

O
AN

G
H

into leading digits (the stem) and

N
AN

C
C

C
the trailing digits (the leaves)
O
C

G
O

N
G
N
 Example

H Data in ordered array:

AN

H
AN

AN
C

21, 24, 24, 26, 27, 27, 30, 32, 38, 41

C
C
O

C
O
G

O
G
N

G
Here, use the 10’s digit for the stem unit:

H
n
N

AN
C
Stem Leaf

C
21 is shown as

O
n 2 1

AN

G
H

N
AN

38 is shown as
C
n 3 8
C

C
O
C

G
O

N
G
N
 Example
(continued)

H Data in ordered array:

AN

H
AN

AN
C

21, 24, 24, 26, 27, 27, 30, 32, 38, 41

C
C
O

C
O
G

O
G
N

H
N

AN
n Completed stem-and-leaf diagram:

C
Stem Leaves

C
H

O
AN
2 1 4 4 6 7 7

G
H

N
AN

C
3 0 2 8
C

C
O
C

4 1
G
O

N
G
N
 Using other stem units

H
AN

H
Using the 100’s digit as the stem:

AN
n

AN
C

C
C

Round off the 10’s digit to form the leaves

O

C
n
O
G

O
G
N

H
N

AN
Stem Leaf

C
n 613 would become 6 1

C
H

O
n 776 would become 7 8

AN

G
H

N
AN

...

C
n
C

C
n 1224 becomes O 12 2
C

G
O

N
G
N
 Using other stem units
(continued)
n Using the 100’s digit as the stem:
H
AN

H
AN

AN
The completed stem-and-leaf display:
C

C
C
O

C
O
G

O
Data:
G
N

H
Stem Leaves
N

AN
6 136
613, 632, 658, 717,

C
C
722, 750, 776, 827, 7 2258

O
841, 859, 863, 891,

AN

G
H

8 346699

N
AN

894, 906, 928, 933,

C
9 13368
C

955, 982, 1034,

C
O
C

1047,1056, 1140, 10 356

G
O

1169, 1224
G

11 47
N

12 2
2.3 Summarising data for two variables using tables and graphics
Relationships Between Variables
Graphs illustrated so far have involved only a
H
n AN

H
single variable

AN

AN
C

C
When two variables exist other techniques are
C

n
O

C
used: O
G

O
G
N

H
N

AN
Categorical Numerical

C
C
(Qualitative) (Quantitative)

O
AN

G
H

Variables Variables

N
AN

C
C

C
O
C

Cross tables Scatter plots

G
O

N
G
N
 2.3.1 Cross Tables

H
AN

H
AN
Cross Tables (or contingency tables) list the

AN
n
C

C
number of observations for every combination
C
O

C
of values for two categorical or ordinal
O
G

O
G
N

H
variables
N

AN
C
C
H

O
If there are r categories for the first variable

AN
n

G
H

N
AN

(rows) and c categories for the second

C
C

C
variable (columns), the table is called an r x c
O
C

G
O

cross table
N
G
N
 Cross Table Example
n 4 x 3 Cross Table for Investment Choices by Investor
H
AN
(values in $1000’s)

H
AN

AN
C

C
C

Investment Investor A Investor B Investor C Total

O

C
Category O
G

O
G
N

H
Stocks 46.5 55 27.5 129
N

AN
C
Bonds 32.0 44 19.0 95

C
H

O
CD 15.5 20 13.5 49

AN

G
H

N
AN

Savings 16.0 28 7.0 51

C
C

C
O
C

Total 110.0 147 67.0 324

G
O

N
G
N
 2.3.2 Charts

H
AN

H
AN

AN
Scatter Diagrams are used for paired
C

C
C

observations taken from two

O

C
O
G

O
G
N

G
numerical variables

H
N

AN
C
C
H

O
n The Scatter Diagram:

AN

G
H

N
AN

n one variable is measured on the vertical

C
C

C
axis and the other variable is measured
O
C

G
O

on the horizontal axis

N
G
N
 Scatter Diagram Example

H
AN

H
AN

AN
C

Volume Cost per Cost per Day vs. Production Volume

C
C

per day day

O

C
O
G

O
23 125 250
G
N

H
N

AN
26 140
200
Cost per Day

29 146

C
150

C
33 160

O
100

AN

G
38 167
H

N
AN

C
42 170 50
C

50 188
C
0
O
C

G
O

55 195
0 10 20 30 40 50 60 70
N
G

60 200
N

Volume per Day

 Graphing
Multivariate Categorical Data
(continued)

H
AN

H
Side by side bar charts

AN

AN
n
C

C
C
O

C
Comparing Investors
O
G

O
G
N

H
N

AN
Savings

C
C
CD

O
AN

G
H

Bonds

N
AN

C
C

Stocks
C
O
C

G
O

N
G

0 10 20 30 40 50 60
N

Investor A Investor B Investor C

 Side-by-Side Chart Example

H
n ANSales by quarter for three sales territories:

H
AN

AN
1st Qtr 2nd Qtr 3rd Qtr 4th Qtr
C

East 20.4 27.4 59 20.4

C
C

West 30.6 38.6 34.6 31.6

O

C
North 45.9 46.9 45 43.9
O
G

O
G
N

H
60
N

AN
C
50

C
H

O
40

AN
East

G
H

N
AN

30 West

C
C

C
O North
20
C

G
O

N
G

10
N

0
1st Qtr 2nd Qtr 3rd Qtr 4th Qtr
 Graphical
Presentation of Data
(continued)
H
AN

H
Techniques reviewed in this chapter:

AN
n

AN
C

C
C
O

C
Categorical
O Numerical
G

O
G
N

H
Variables Variables
N

AN
C
C
• Frequency distribution • Line chart

O
AN

G
• Bar chart • Frequency distribution
H

N
AN

C
• Pie chart • Histogram and ogive
C

• Pareto diagram C • Stem-and-leaf display

O
C

G
O

• Scatter plot
N
G
N
2.4 Summarizing data for a quantitative variable
using numerical measures

H
Describing Data Numerically

AN

H
C

AN
C

H
O

C
Central Tendency Variation

AN
G

C
N

C
G

C
Arithmetic Mean Range

O
G
N
Median Interquartile Range
H
AN

Mode Variance
C

H
AN
C

H
Standard Deviation
O

AN
C
G
N

C
Coefficient of Variation
O

C
G

O
N

G
N
 2.4.1 Measures of center location
(measures of central tendency)

H
Overview
AN

H
C Central Tendency

AN
C

H
O

AN
G

C
N

C
G

C
N

O
Mean Median Mode

G
N
H
AN

åx
C

H
i
AN
C

x= i=1

H
O

AN
n
C
G
N

C
O

Arithmetic Midpoint of Most frequently

C
G

O
average ranked values observed value
N

G
N
 Arithmetic Mean
n The arithmetic mean (mean) is the most

H
AN
common measure of central tendency

H
C

AN
C
For a population of N values:

H
O

C
n

AN
G

C
N

C
åx

C
x1 + x 2 +  + x N
i

N
Population

O
μ= =
i=1

G
values

N
N N
H

Population size
AN
C

n For a sample of size n:

H
AN
C

H
O

åx

AN
C
G

i
x1 + x 2 +  + x n Observed
N

C
x= i=1
=
O

values

C
G

n n

O
N

G
Sample size

N
 Arithmetic Mean

H
(continued)

AN

H
C

AN
n The most common measure of central tendency
C

H
O

C
Mean = sum of values divided by the number of values

AN
G

C
N

C
Affected by extreme values (outliers)

G
n

C
N

O
G
N
H

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
AN
C

H
AN
C

H
O

AN
Mean = 3 Mean = 4
C
G
N

C
O

1 + 2 + 3 + 4 + 5 15 1 + 2 + 3 + 4 + 10 20

C
G

= =3 = =4

O
N

G
5 5 5 5

N
 Median

H
AN

H
n In an ordered list, the median is the “middle”
C

AN
C

H
number (50% above, 50% below)
O

AN
G

C
N

C
G

C
N

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

G
N
H
AN

Median = 3 Median = 3
C

H
AN
C

H
O

AN
C
G
N

C
Not affected by extreme values
O

C
G

O
N

G
N
 Finding the Median

H
AN
n The location of the median:

H
C

AN
C

n +1

H
O

C
Median position = position in the ordered data

AN
G

C
2
N

C
G

C
N

O
n If the number of values is odd, the median is the middle number

G
N
n If the number of values is even, the median is the average of
H

the two middle numbers

AN
C

n +1 H
AN
C

H
O

Note that is not the value of the median, only the

AN
n
C
G

2
N

C
position of the median in the ranked data
O

C
G

O
N

G
N
 Mode
n A measure of central tendency

H
AN
Value that occurs most often

H
n C

AN
C

Not affected by extreme values

H
O

C
n

AN
G

C
N

Used for either numerical or categorical data

C
n

C
N

O
n There may may be no mode

G
N
There may be several modes
H

n
AN
C

H
AN
C

H
O

AN
C
G
N

C
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
O

C
G

0 1No
2 Mode
3 4 5 6

O
N

Mode = 9

G
N
 Review Example:
Summary Statistics

H
AN
n Five houses on a hill by the beach

H
C

AN
C
House Prices:

H
Mean: ($3,000,000/5)
O

C
n

AN
G

C
N

$2,000,000

O
= $600,000

C
G
500,000

C
N

O
300,000

G
N
100,000
Median: middle value of ranked data
H

n
100,000
AN

= $300,000
C

H
Sum 3,000,000
AN
C

H
O

AN
C
G

Mode: most frequent value

N

C
n
O

C
= $100,000
G

O
N

G
N
 Which measure of location
is the “best”?

H
AN

H
C

AN
n Mean is generally used, unless extreme
C

H
O

C
values (outliers) exist . . .

AN
G

C
N

C
G

C
Then median is often used, since the median

N
n

O
G
is not sensitive to extreme values.

N
H
AN

n Example: Median home prices may be reported for

C

a region – less sensitive to outliers

H
AN
C

H
O

AN
C
G
N

C
O

C
G

O
N

G
N
 Shape of a Distribution

H
Describes how data are distributed
AN
n

H
C
Measures of shape

AN
n
C

H
O

AN
Symmetric or skewed
G

C
n
N

C
G

C
N

O
G
N
H

Left-Skewed Symmetric Right-Skewed

AN
C

Mean < Median

H
Mean = Median Median < Mean
AN
C

H
O

AN
C
G
N

C
O

C
G

O
N

G
N
 2.4.2 Measures of dispersion
(measures of variability)
Variation

H
AN

H
C

AN
C
Range Interquartile Variance Standard Coefficient of

H
O

AN
G

Range Deviation Variation

C
N

C
G

C
N

O
G
N
n Measures of variation give
H
AN

information on the spread

C

or variability of the data

H
AN
C

H
values.
O

AN
C
G
N

C
O

Same center,

C
G

O
N

different variation

G
N
 Range

H
AN
n Simplest measure of variation

H
C

AN
Difference between the largest and the smallest
C
n

H
O

AN
observations:
G

C
N

C
G

C
N
Range = Xlargest – Xsmallest

O
G
N
H
AN

Example:
C

H
AN
C

H
O

AN
C

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
G
N

C
O

C
Range = 14 - 1 = 13
G

O
N

G
N
 Disadvantages of the Range
Ignores the way in which data are distributed

H
n

AN

H
C

AN
C

H
O

C
7 8 9 10 11 12 7 8 9 10 11 12

AN
G

C
N

Range = 12 - 7 = 5 Range = 12 - 7 = 5

C
G

C
N

O
Sensitive to outliers

G
n

N
H
AN

1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,3,3,3,3,4,5
C

H
Range = 5 - 1 = 4
AN
C

H
O

AN
C
G
N

1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,3,3,3,3,4,120
C

C
O

C
G

Range = 120 - 1 = 119

O
N

G
N
 Interquartile Range

H
Can eliminate some outlier problems by using
AN
n

H
the interquartile range
C

AN
C

H
O

AN
G

C
N

Eliminate high- and low-valued observations

C
n

C
and calculate the range of the middle 50% of

O
G
the data

N
H
AN
C

n H
Interquartile range = 3rd quartile – 1st quartile
AN
C

H
O

AN
IQR = Q3 – Q1
C
G
N

C
O

C
G

O
N

G
N
 Interquartile Range

H
AN
Example:

H
C

AN
Median
C
X X

H
Q1 Q3
O

C
minimum (Q2) maximum

AN
G

C
N

C
25% 25% 25% 25%

C
N

O
G
12 30 45 57 70

N
H
AN
C

H
Interquartile range
AN
C

H
O

= 57 – 30 = 27

AN
C
G
N

C
O

C
G

O
N

G
N
 Quartiles

H
Quartiles split the ranked data into 4 segments with
AN
n

H
an equal number of values per segment
C

AN
C

H
O

AN
G

C
N

C
25% 25% 25% 25%

C
N

O
G
Q1 Q2 Q3

N
H
AN

n The first quartile, Q1, is the value for which 25% of the
C

H
observations are smaller and 75% are larger
AN
C

H
O

Q2 is the same as the median (50% are smaller, 50% are

AN
C

n
G

larger)
N

C
O

C
Only 25% of the observations are greater than the third
G

O
N

quartile

G
N
 Quartile Formulas

H
AN

H
C

AN
Find a quartile by determining the value in the
C

H
O

AN
appropriate position in the ranked data, where
G

C
N

C
G

C
N

O
First quartile position: Q1 = 0.25(n+1)

G
N
H
AN

Second quartile position: Q2 = 0.50(n+1)

C

(the median position)

H
AN
C

H
O

AN
C
G

Third quartile position: Q3 = 0.75(n+1)

N

C
O

C
G

O
N

G
where n is the number of observed values

N
 Quartiles

H
AN

H
C
Example: Find the first quartile

AN
n
C

H
O

AN
G

C
N

Sample Ranked Data: 11 12 13 16 16 17 18 21 22

C
G

C
N

O
(n = 9)

G
N
H

Q1 = is in the 0.25(9+1) = 2.5 position of the ranked data

AN
C

so use the value half way between the 2nd and 3rd values,
H
AN
C

H
O

AN
C
G

so Q1 = 12.5
N

C
O

C
G

O
N

G
N
 Population Variance

H
Average of squared deviations of values from
AN
n

H
C
the mean

AN
C

H
O

AN
G

C
N
N

O
å (x - μ)

C
Population variance: 2

G
n

C
N
i

O
G
σ =2 i=1

N
H

N
AN
C

H
AN
C

Where μ = population mean

H
O

AN
C
G

N = population size
N

C
O

C
G

xi = ith value of the variable x

O
N

G
N
 Sample Variance

H
Average (approximately) of squared deviations
AN
n

H
C
of values from the mean

AN
C

H
O

AN
G

C
n
N

O
å (x - x)

C
Sample variance: 2

G
n

C
N
i

O
G
s =
2 i=1

N
H

n -1
AN
C

H
AN
C

Where X = arithmetic mean

H
O

AN
C
G

n = sample size
N

C
O

C
G

Xi = ith value of the variable X

O
N

G
N
 Population Standard Deviation

H
Most commonly used measure of variation
AN
n

H
C
Shows variation about the mean

AN
n
C

H
O

AN
Has the same units as the original data
G

C
n
N

C
G

C
N

O
G
n Population standard deviation:

N
H
AN

å (x - μ)
C

H 2
AN
C

H
i
O

AN
C
G

σ= i=1
N

C
N
O

C
G

O
N

G
N
 Sample Standard Deviation

H
AN

H
n Most commonly used measure of variation
C

AN
C

H
O

C
Shows variation about the mean

AN
n
G

C
N

C
Has the same units as the original data

G
n

C
N

O
G
N
H

Sample standard deviation: n

å i
AN

n
(x - x) 2
C

H
AN
S=
C

i=1

H
O

AN
C

n -1
G
N

C
O

C
G

O
N

G
N
 Calculation Example:
Sample Standard Deviation

H
AN

H
Sample C

AN
C
Data (xi) : 10 12 14 15 17 18 18 24

H
O

AN
G

C
N

n=8 Mean = x = 16

C
G

C
N

O
(10 - X)2 + (12 - x)2 + (14 - x)2 +  + (24 - x)2

G
s=

N
n -1
H
AN
C

H
AN
C

(10 - 16)2 + (12 - 16)2 + (14 - 16)2 +  + (24 - 16)2

H
O

AN
C
G

8 -1
N

C
O

C
G

O
N

A measure of the “average”

G
126
= = 4.2426

N
7 scatter around the mean
 Measuring variation

H
AN

H
C

AN
C

H
O

AN
G

Small standard deviation

C
N

C
G

C
N

O
G
Large standard deviation

N
H
AN
C

H
AN
C

H
O

AN
C
G
N

C
O

C
G

O
N

G
N
Comparing Standard Deviations

H
AN

H
C

AN
Data A
C

H
O

Mean = 15.5

AN
G

C
s = 3.338
N

C
11 12 13 14 15 16 17 18 19 20 21

C
N

O
G
Data B

N
H

Mean = 15.5
AN

11 12 13 14 15 16 17 18 19 20 21 s = 0.926
C

H
AN
C

H
O

AN
C

Data C
G
N

C
Mean = 15.5
O

C
G

s = 4.570

O
N

11 12 13 14 15 16 17 18 19 20 21

G
N
 Advantages of Variance and
Standard Deviation

H
AN

H
C

AN
C

H
O

C
Each value in the data set is used in the

AN
n
G

C
N

O
calculation

C
G

C
N

O
G
N
H

n Values far from the mean are given extra

AN

weight
C

H
AN
C

H
(because deviations from the mean are squared)
O

AN
C
G
N

C
O

C
G

O
N

G
N
 Coefficient of Variation

H
Measures relative variation
AN
n

H
C
Always in percentage (%)

AN
n
C

H
O

AN
Shows variation relative to mean
G

C
n
N

C
G
Can be used to compare two or more sets of

C
N
n

O
G
data measured in different units

N
H
AN
C

æ s öH
AN
C

CV = çç ÷÷ × 100%

H
O

AN
C
G

èx ø
N

C
O

C
G

O
N

G
N
 Comparing Coefficient
of Variation

H
AN
Stock A:

H
n C

AN
C
n Average price last year = $50

H
O

AN
G

C
n Standard deviation = $5
N

C
G
æs ö

C
$5

O
CVA = çç ÷÷ × 100% = × 100% = 10%

G
èx ø $50

N
Both stocks
H

Stock B: have the same

AN

n
standard
C

n
H
Average price last year = $100 deviation, but
AN
C

H
stock B is less
O

Standard deviation = $5

AN
C
G

n
variable relative
N

C
to its price
O

æs ö $5

C
G

CVB = çç ÷÷ × 100% = × 100% = 5%

O
N

G
èx ø $100

N
 Chebychev’s Theorem

H
AN

H
C
For any population with mean μ and

AN
n
C

H
O

C
standard deviation σ , and k > 1 , the

AN
G

C
N

percentage of observations that fall within

C
G

C
N
the interval

O
G
N
[μ + kσ]
H
AN
C

Is at least AN
H
C

H
O

AN
C
G
N

100[1 - (1/k )]%

2

C
O

C
G

O
N

G
N
 Chebychev’s Theorem
(continued)

H
AN
Regardless of how the data are distributed, at

H
n C

AN
C

least (1 - 1/k2) of the values will fall within k

H
O

AN
G

C
standard deviations of the mean (for k > 1)
N

C
G

C
N

O
n Examples:

G
N
H
AN

At least within
C

H
AN
C

H
(1 - 1/1.52) = 55.6% ……... k = 1.5 (μ ± 1.5σ)
O

AN
C
G

(1 - 1/22) = 75% …........... k = 2 (μ ± 2σ)

N

C
O

C
G

(1 - 1/32) = 89% …….…... k = 3 (μ ± 3σ)

O
N

G
N
 The Empirical Rule

H
AN

H
If the data distribution is bell-shaped, then
C

AN
n
C

H
O

C
the interval:

AN
G

C
N

C
μ ± 1σ contains about 68% of the values in

C
n

O
G
the population or the sample

N
H
AN
C

H
AN
C

H
O

68%

AN
C
G
N

C
O

C
μ
G

O
N

G
μ ± 1σ

N
 The Empirical Rule

H
AN
μ ± 2σ contains about 95% of the values in

H
n C

AN
C

the population or the sample

H
O

AN
G

C
μ ± 3σ contains almost all (about 99.7%) of
N

C
n

C
the values in the population or the sample

O
G
N
H
AN
C

H
AN
C

H
O

95% 99.7%

AN
C
G
N

C
O

C
G

O
μ ± 2σ μ ± 3σ
N

G
N
 2.4.3 Weighted Mean

H
n The weighted mean of a set of data is
AN

H
C

AN
C

H
O

AN
G

C
n
N

åw x

C
G

C
i i
w 1x1 + w 2 x 2 +  + w n x n

O
x= i=1
=

G
Where wi nis the weight of the ith observation
n

N
n
H
AN

and
C

n = å wi H
AN
C

H
O

AN
Use when data is already grouped into n classes, with
C
G

n
N

wi values in the ith class

C
O

C
G

O
N

G
N
Approximations for Grouped Data

H
AN
Suppose data are grouped into K classes, with

H
C

AN
frequencies f1, f2, . . . fK, and the midpoints of the
C

H
O

C
classes are m1, m2, . . ., mK

AN
G

C
N

C
G

C
N

O
G
n For a sample of n observations, the mean is

N
H
AN

åfm
C

H where n = å fi
i i
AN
C

x= i=1

H
O

i=1

AN
C

n
G
N

C
O

C
G

O
N

G
N
 Approximations for Grouped Data

H
AN
Suppose data are grouped into K classes, with

H
C

AN
frequencies f1, f2, . . . fK, and the midpoints of the
C

H
O

C
classes are m1, m2, . . ., mK

AN
G

C
N

C
G

C
N

O
G
n For a sample of n observations, the variance is

N
H
AN

åi i
f (m - x) 2
C

H
AN
C

s2 =

H
i=1
O

AN
C

n -1
G
N

C
O

C
G

O
N

G
N
 2.6 Measures of association between two
quantitative variables

H
The covariance measures the strength of the linear relationship

AN
n
between two variables

H
C

AN
C

H
The population covariance:
O

C
n

AN
G

C
N

C
å (x - µ )(y i - µ y )

C
i x

O
Cov (x , y) = s xy = i=1

G
N

N
H
AN

n The sample covariance:

C

H n
AN
C

å (x - x)(y - y)

H
O

AN
C

i i
G

Cov (x , y) = s xy = i=1
N

C
n -1
O

C
G

O
Only concerned with the strength of the relationship
N

G
N
n No causal effect is implied
 Interpreting Covariance

H
AN

H
C

AN
C

Covariance between two variables:

H
O

C
n

AN
G

C
N

C
G

C
N

O
G
Cov(x,y) > 0 x and y tend to move in the same direction

N
H
AN

Cov(x,y) < 0 x and y tend to move in opposite directions

C

H
AN
C

H
Cov(x,y) = 0 x and y are independent
O

AN
C
G
N

C
O

C
G

O
N

G
N
 Coefficient of Correlation

H
AN
Measures the relative strength of the linear relationship

H
n C

AN
between two variables
C

H
O

AN
G

C
N

O
Population correlation coefficient:

C
n

C
N

O
Cov (x , y)

G
ρ=

N
H

σXσY
AN
C

H
AN
C

Sample correlation coefficient:

H
n
O

AN
C
G

Cov (x , y)
N

C
r=
O

C
G

sX sY

O
N

G
N
 Features of
Correlation Coefficient, r

H
AN

H
C
Unit free

AN
n
C

H
O

AN
Ranges between –1 and 1
G

C
N

C
G
The closer to –1, the stronger the negative linear

C
n

O
relationship

G
N
H

The closer to 1, the stronger the positive linear

AN

n
C

relationship
H
AN
C

H
O

The closer to 0, the weaker any positive linear

AN
C

n
G
N

C
relationship
O

C
G

O
N

G
N
 Scatter Plots of Data with Various
Correlation Coefficients

H
AN
Y Y Y

H
C

AN
C

H
O

AN
G

C
N

C
G

C
N

O
X X X

G
N
r = -1 r = -.6 r=0
H
AN

Y
C

Y Y H
AN
C

H
O

AN
C
G
N

C
O

C
G

O
N

G
X X X

N
r = +1 r = +.3 r=0
 Chapter Summary
n Reviewed types of data and measurement levels
H

H
H
AN

AN
Data in raw form are usually not easy to use for decision

AN
n
C

C
making -- Some type of organization is needed:

C
C

C
C
O

O
¨ Table ¨ Graph

O
G

G
G
N

N
N
n Techniques reviewed in this chapter:
H
AN

H
Frequency distribution Line chart
AN
n n
C

Frequency distribution

H
Bar chart n
C

C
n

AN
O

Histogram and ogive

C

n
Pie chart
G

C
Stem-and-leaf display
N

n
Pareto diagram

C
n
N

Scatter plot
O
n

n G
Cross tables and
N
side-by-side bar charts
 Chapter Summary
(continued)

H
n Described measures of central tendency
AN

H
n Mean, median, mode
C

AN
C

Illustrated the shape of the distribution

H
O

C
n

AN
G

C
N

Symmetric, skewed

C
n

C
Described measures of variation

O
n

G
Range, interquartile range, variance and standard deviation,

N
n
H

coefficient of variation
AN

Discussed measures of grouped data

C

H
n
AN
C

H
O

Calculated measures of relationships between

AN
C

n
G
N

C
variables
O

C
G

O
covariance and correlation coefficient
N

G
N