Data Types, Data Display, and Summary Statistics

Professor Turetsky
Deck 1
Types of Data

2
Types of Data

• Univariate or Multivariate?
– Univariate: one fact for each object in a dataset (“one column in a spreadsheet”)
– Multivariate: two or more facts for each object in a dataset (“many columns in a spreadsheet”)

3
Types of Data

• Quantitative or Qualitative?
– Quantitative: presented as numbers permitting arithmetic
• Interest rate
• Temperature
– Qualitative (categorical): everything else
• Country of birth
• Supplier/Vendor

4
Types of Data (Quantitative)

• Discrete or Continuous?
– Discrete: counted
• Cars sold
• Number of children
– Continuous: measured (always allow “in-between” values)
• Gallons of oil sold
• Temperature

– What about age? Money?

5
Types of Data (Qualitative)

• Nominal Data
– Definition: “Qualitative data that has no ordering”
– Example:

6
Types of Data (Qualitative)

• Ordinal Data
– Definition: “Qualitative data that has an ordering”
– Example – Likert Scale:

disagree strongly  disagree  neutral  agree  agree strongly

– Often “measure” with numbers:
1 = disagree strongly
2 = disagree

5 = agree strongly

7
Types of Data (Time Series)

• Time Series or Cross-Sectional?

– Time series: when time sequencing is important
• US historical inflation rates
• A baby’s weight
– Cross-sectional: data are contemporaneous, or time doesn’t matter
• Inflation rates for several countries
• Weight at birth

8
Population vs. Sample

• Population
– Entire Set of Observations
• The entire population is assumed to be fixed.
• Often hypothetical since we generally cannot collect the entire population
• However, must be defined
• Sample
– A subset of observations from the population
• Easier to collect
• Sample needs to be representative of the population.
• Ultimately, we want to infer something about the population.

9
Random Sample

• Random Sample:
– All elements of the population have the same chance of being chosen in the sample.
• Hence, our sample will be representative of the population.
• There are various types of random sampling procedures.

10
Population vs Sample

• Population is Fixed
• Sample is Random
• Parameter: Numerical Summary Measure of Population
• Statistic: Numerical Summary Measure of Sample
• Parameters are fixed while Statistics are RANDOM.

11
Estimator vs. Estimate

• Estimator:
– A statistical method or formula used to calculate an estimate for a population parameter
• Estimate:
– The specific value that is calculated using an estimator on the sample data

12
Descriptive Statistics:
Graphical Summaries

13
Data Display

• In an MBA class, we know about each student:

– First name
– Last name
– Age
– Undergraduate major
– Work experience (in years)
– Annual Compensation

14
Dataset

15
The Bar Chart

Horizontal axis is qualitative

16
The Stacked Bar Chart

17
The Histogram

• Horizontal axis is quantitative

• No space between columns
18
The Histogram
• With a histogram, can detect
– Skewness:

– Outliers:

– Bimodal distribution:

19
A Bit More on Skewness
The Scatter Plot

21
The Scatter Plot
• Positive (linear) relation:

• Negative (linear) relation:

• No relation:

22
Time Series

23
Descriptive Statistics:
Numerical Summaries

24
The Sample Mean
• Compensation Data from MBA students: Category - Other
– n = 9 numbers:
86.24, 94.62, 95.64, 95.60, 95.42, 88.77, 88.95, 99.80, 97.18

x1 x2 x3 x4 x5 x6 x7 x8 xn

• Sample Mean: 93.58

n
x 1
n x
i 1
i  1
n x
i
i  1
n x

25
Population Mean

26
The Median

• Median = the middle value in a sorted dataset

86.24, 88.77, 88.95, 94.62, 95.42, 95.60, 95.64, 97.18, 99.80

• When n is odd, take “true” middle value

• When n is even, take average of two middle values

27
Mean vs. Median

• Mean is
– Sensitive to outliers
– Includes “more information”
• Median is
– Useful when ranking is important
– Important in demographics
• Other “typical value” measures
– Mode = the most common value
– Trimmed mean (ignore upper and lower x% of data)

28
Variability: Is it Good or Bad?

Histogram of Price
0 1
40

30
Frequency

20

10

0
20 30 40 50 60 70 20 30 40 50 60 70
Price
Panel variable: Before/After

29
Measuring Variability

30
Population Variance and Standard Deviation

31
Sample Variance: Calculating s2
x
(xi - x)
 2
s 
2 1
n-1   (x i - x) 
 i 

 
s2  n1-1   x i  nx 2 
2

 i 

32
Sample Variance vs Population Variance

 
s2  n11   (x i - x)2 
 i 

33
Using Mean and Standard Deviation

• Empirical Rule (assuming distribution is somewhat bell shaped and symmetric)

• Approximately 68% of observations are within 1 standard deviation of the mean
• Approximately 95% of observations are within 2 standard deviation of the mean
• Approximately 99% of observations are within 3 standard deviation of the mean

34
Z-transformation

35
Quartiles

• First quartile = the value in position (n+1)/4 in a sorted dataset

– Also called QL, 25th percentile, Q1
• Median is the second quartile
• Third quartile = the value in position 3(n+1)/4 in a sorted dataset
– Also called QU, 75th percentile, Q3

36
Measures of Variability

• Range = max – min (very sensitive to outliers)

• σ ≡ population standard deviation and s ≡ sample standard deviation (sensitive to
outliers, why?)
• IQR = QU – QL (not sensitive to outliers)
– Can be used to detect outliers.

37
The Box Plot

• Inter-Quartile Range (IQR) = QU – QL

• Inner fences:
– Lower inner fence = QL – 1.5 IQR
– Upper inner fence = QU + 1.5 IQR
• Outer fences:
– Lower outer fence = QL – 3.0 IQR
– Upper outer fence = QU + 3.0 IQR
• Data outside inner fence = outlier
• Data outside outer fence = serious outlier

38
The Box Plot
• Box edges (“hinges”) at QU and QL; mark median with bar
• Draw “whiskers” from edges to smallest and largest non-
outliers

39
The Box Plot

40
The Box Plot

41
Transformations

• Often we want to convert a skewed dataset to a more symmetric one

• Can use the Log transformation

42