CHAPTER 3

Numerical Descriptive Measures

USING STATISTICS: Evaluating the Performance of Mutual Funds
3.1

3.2

MEASURES OF CENTRAL TENDENCY,

VARIATION, AND SHAPE
The Mean
The Median
The Mode
Quartiles
The Geometric Mean
The Range
The Interquartile Range
The Variance and the Standard Deviation
The Coefficient of Variation
Z Scores
Shape
Visual Explorations: Exploring Descriptive
Statistics
Microsoft Excel Descriptive Statistics Output
Minitab Descriptive Statistics Output
NUMERICAL DESCRIPTIVE MEASURES
FOR A POPULATION
The Population Mean
The Population Variance and Standard Deviation

The Empirical Rule

The Chebychev Rule
3.3

COMPUTING NUMERICAL DESCRIPTIVE

MEASURES FROM A FREQUENCY
DISTRIBUTION

3.4

EXPLORATORY DATA ANALYSIS

The Five-Number Summary
The Box-and-Whisker Plot

3.5

THE COVARIANCE AND THE COEFFICIENT

OF CORRELATION
The Covariance
The Coefficient of Correlation

3.6

PITFALLS IN NUMERICAL DESCRIPTIVE

MEASURES AND ETHICAL ISSUES

A.3

USING SOFTWARE FOR DESCRIPTIVE

STATISTICS
A.3.1 Microsoft Excel
A3.2 Minitab
A3.3 (CD-ROM Topic) SPSS

LEARNING OBJECTIVES
In this chapter, you learn:
To describe the properties of central tendency, variation,
and shape in numerical data
To calculate descriptive summary measures for a population
To construct and interpret a box-and-whisker plot
To describe the covariance and the coefficient of correlation

72

CHAPTER THREE Numerical Descriptive Measures

U S I N G S TAT I S T I C S
Evaluating the Performance of Mutual Funds
Return to the study of mutual funds introduced in Chapter 2. You want to
decide which types of mutual funds to invest in. In the last chapter you
learned how to present data in tables and charts. However, when dealing
with numerical data, such as the return on investments in mutual funds in
2003, you also need to summarize the data, and ask statistical questions.
What is the central tendency for returns of the various funds? For example, what is the mean return in 2003 for the low-risk, average-risk, and
high-risk mutual funds? How much variability is present in the returns?
Are the returns for high-risk funds more variable than for average-risk
funds or low-risk funds? How can you use this information when deciding
what mutual funds to invest in?

or numerical variables, you need more than just the visual picture of what a variable looks
like than you get from the graphs discussed in Chapter 2. For example, for the 2003 returns,
you would like to determine not only whether the riskier funds had a higher 2003 return, but
whether they also had greater variation, and how the returns for each risk group were distributed. You also want to examine whether there is a relationship between the expense ratio and
the 2003 return. Reading this chapter will allow you to learn about some of the methods to
measure:

central tendency, the extent to which all of the data values group around a central value
variation, the amount of dispersion or scattering of values away from a central point
shape, the pattern of the distribution of values from the lowest value to the highest value

You will also learn about the covariance and the coefficient of correlation that help measure the
strength of the association between two numerical variables.

3.1

MEASURES OF CENTRAL TENDENCY, VARIATION,

AND SHAPE
You can characterize any set of data by measuring its central tendency, variation, and shape.
Most sets of data show a distinct central tendency to group around a central point. When people talk about an average value or the middle value or the most popular or frequent value,
they are talking informally about the mean, median, and mode, three measures of central tendency.
Variation measures the spread or dispersion of values in a data set. One simple measure of
variation is the range, the difference between the highest and lowest value. More commonly
used in statistics are the standard deviation and variance, two measures explained later in this
section. The shape of a data set represents a pattern of all the values from the lowest to highest
value. As you will learn later in this section, many data sets have a pattern that looks approximately like a bell, with a peak of values somewhere in the middle.

3.1: Measures of Central Tendency, Variation, and Shape

73

The Mean
The arithmetic mean (typically referred to as the mean) is the most common measure of central tendency. The mean is the only common measure in which all the values play an equal role.
The mean serves as a balance point in a set of data (like the fulcrum on a seesaw). You calculate the mean by adding together all the values in a data set and then dividing that sum by the
number of values in the data set.
The symbol X , called X bar, is used to represent the mean of a sample. For a sample containing n values, the equation for the mean of a sample, is written as
sum of the values
number of values

X =

Using the series X1, X2, . . . , Xn to represent the set of n values and n to represent the number of
values, the equation becomes:
X =

X1 + X 2 + L + X n
n

By using summation notation (discussed fully in Appendix B), you replace the numerator
n

X 1 + X 2 + + X n by the term

Xi

that means sum all the X i values from the first X

i =1

value, X1 , to the last X value, Xn , to form Equation (3.1), a formal definition of the sample
mean.
SAMPLE MEAN
The sample mean is the sum of the values divided by the number of values.
n

X =

Xi
i =1

(3.1)

X = sample mean
n = number of values or sample size

where

Xi = ith value of the variable X

X i = summation of all Xi values in the sample

i =1

Because all the values play an equal role, a mean will be greatly affected by any value that
is greatly different from the others in the data set. When you have such extreme values, you
should avoid using the mean.
The mean can suggest what is a typical or central value for a data set. For example, if you
knew the typical time it takes you to get ready in the morning, you might be able to better plan
your morning and minimize any excessive lateness (or earliness) going to your destination.
Suppose you define the time to get ready as the time in minutes (rounded to the nearest minute)
from when you get out of bed to when you leave your home. You collect the times shown below
for 10 consecutive work days:
Day:
Time (minutes):

10

39

29

43

52

39

44

40

31

44

35

74

CHAPTER THREE Numerical Descriptive Measures

TIMES

The mean time is 39.6 minutes, computed as follows:

X =

sum of the values

number of values
n

X =

Xi
i =1

X =

39 + 29 + 43 + 52 + 39 + 44 + 40 + 31 + 44 + 35
10

X =

396
= 39.6
10

Even though no one day in the sample actually had the value 39.6 minutes, allotting about 40
minutes to get ready would be a good rule for planning your mornings, but only because the 10
days does not contain extreme values.
Contrast this to the case in which the value on day four was 102 minutes instead of 52 minutes. This extreme value would cause the mean to rise to 44.6 minutes as follows:
X =

sum of the values

number of values
n

X =
X =

Xi
i =1

n
446
= 44.6
10

The one extreme value has increased the mean by more than 10% from 39.6 to 44.6 minutes. In
contrast to the original mean that was in the middle, greater than 5 of the get-ready times
(and less than the 5 other times), the new mean is greater than 9 of the 10 get-ready times. The
extreme value has caused the mean to be a poor measure of central tendency.
EXAMPLE 3.1

THE MEAN 2003 RETURN FOR SMALL CAP MUTUAL FUNDS WITH HIGH RISK
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified
according to the risk level of the mutual funds (low, average, and high) and type (small cap, mid
cap, and large cap). Compute the mean 2003 return for the small cap mutual funds with high risk.
SOLUTION The mean 2003 return for the small cap mutual funds with high risk (MUTUALis 51.53, calculated as follows:

FUNDS2004)

X =

sum of the values

number of values
n

=
=

Xi
i =1

n
463.8
= 51.53
9

The ordered array for the nine small cap mutual funds with high risk is:
37.3

39.2

44.2

44.5

53.8

56.6

59.3

62.4

66.5

Four of these returns are below the mean of 51.53 and five of these returns are above the mean.

3.1: Measures of Central Tendency, Variation, and Shape

75

The Median
The median is the value that splits a ranked set of data into two equal parts. The median is not
affected by extreme values, so you can use the median when extreme values are present.
The median is the middle value in a set of data that has been ordered from lowest to highest
value.
To calculate the median for a set of data, you first rank the values from smallest to largest.
Then use Equation (3.2) to compute the rank of the value that is the median.
MEDIAN
50% of the values are smaller than the median and 50% of the values are larger than the
median.
Median =

n +1
ranked value
2

(3.2)

You compute the median value by following one of two rules:

Rule 1 If there are an odd number of values in the data set, the median is the middle
ranked value.
Rule 2 If there are an even number of values in the data set, then the median is the average
of the two middle ranked values.

To compute the median for the sample of 10 times to get ready in the morning, you rank the
daily times as follows:
Ranked values:
29 31 35 39 39 40 43 44 44 52
Ranks:
1

9 10

Median = 39.5
Because the result of dividing n + 1 by 2 is (10 + 1)/2 = 5.5 for this sample of 10, you must use
Rule 2 and average the fifth and sixth ranked values, 39 and 40. Therefore, the median is 39.5.
The median of 39.5 means that for half of the days, the time to get ready is less than or equal to
39.5 minutes, and for half of the days the time to get ready is greater than or equal to 39.5 minutes. The median time to get ready of 39.5 minutes is very close to the mean time to get ready
of 39.6 minutes.

EXAMPLE 3.2

COMPUTING THE MEDIAN FROM AN ODD-SIZED SAMPLE

The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the median 2003 return for the nine small cap mutual
funds with high risk. MUTUALFUNDS2004
SOLUTION Because the result of dividing n + 1 by 2 is (9 + 1)/2 = 5 for this sample of nine,
using Rule 1, the median is the fifth ranked value. The percentage return in 2003 for the nine
small cap mutual funds with high risk are ranked from the smallest to the largest:

76

CHAPTER THREE Numerical Descriptive Measures

Ranked values:
37.3

39.2

44.2

44.5

53.8

56.6

59.3

62.4

66.5

Ranks:

Median
The median return is 53.8. Half the small cap high-risk mutual funds have returns equal to or
below 53.8 and half have returns equal to or above 53.8.

The Mode
The mode is the value in a set of data that appears most frequently. Like the median and unlike
the mean, extreme values do not affect the mode. You should use the mode only for descriptive
purposes as it is more variable from sample to sample than either the mean or the median.
Often there is no mode or there are several modes in a set of data. For example, consider the
time to get ready data shown below.
29

31

35

39

39

40

43

44

44

52

There are two modes, 39 minutes and 44 minutes, since each of these values occurs twice.

EXAMPLE 3.3

COMPUTING THE MODE

A systems manager in charge of a companys network keeps track of the number of server failures that occur in a day. Compute the mode for the following data that represents the number of
server failures in a day for the past two weeks.
1 3 0 3 26 2 7 4 0 2 3 3 6 3
SOLUTION The ordered array for these data is
0 0 1 2 2 3 3 3 3 3 4 6 7 26
Because 3 appears five times, more times than any other value, the mode is 3. Thus, the systems
manager can say that the most common occurrence is having three server failures in a day. For this
data set, the median is also equal to 3 while the mean is equal to 4.5. The extreme value 26 is an
outlier. For these data, the median and the mode better measure central tendency than the mean.

A set of data will have no mode if none of the values is most typical. Example 3.4 presents a
data set with no mode.

EXAMPLE 3.4

DATA WITH NO MODE

Compute the mode for the 2003 return for the small cap mutual funds with high risk. MUTUALFUNDS2004

SOLUTION The ordered array for these data is

37.3 39.2 44.2 44.5 53.8 56.6 59.3 62.4 66.5
These data have no mode. None of the values is most typical because each value appears once.

3.1: Measures of Central Tendency, Variation, and Shape

77

Quartiles

1The

Q1, median, and Q3

are also the 25th, 50th, and
75th percentile,
respectively. Equations (3.2),
(3.3), and (3.4) can be
expressed generally in
terms of finding percentiles:
(p 100)th percentile =
p (n + 1) ranked value.

Quartiles split a set of data into four equal partsthe first quartile Q1 divides the smallest
25.0% of the values from the other 75.0% that are larger. The second quartile Q2 is the
median50.0% of the values are smaller than the median and 50.0% are larger. The third
quartile Q3 divides the smallest 75.0% of the values from the largest 25.0%. Equations (3.3)
and (3.4) define the first and third quartiles.1
FIRST QUARTILE Q1
25.0% of the values are smaller than Q1, the first quartile, and 75.0% are larger than the
first quartile Q1.
Q1 =

n +1
ranked value
4

(3.3)

THIRD QUARTILE Q3
75.0% of the values are smaller than the third quartile Q3, and 25.0% are larger than the
third quartile Q3.
Q3 =

3( n + 1)
ranked value
4

(3.4)

Use the following rules to calculate the quartiles:

Rule 1 If the result is a whole number, then the quartile is equal to that ranked value. For example, if the sample size n = 7, the first quartile Q1 is equal to the (7 + 1)/4 = second ranked value.
Rule 2 If the result is a fractional half (2.5, 4.5, etc.), then the quartile is equal to the average of the corresponding ranked values. For example, if the sample size n = 9, the first
quartile Q1 is equal to the (9 + 1)/4 = 2.5 ranked value, halfway between the second ranked
value and the third ranked value.
Rule 3 If the result is neither a whole number nor a fractional half, you round the result to
the nearest integer and select that ranked value. For example, if the sample size n = 10, the
first quartile Q1 is equal to the (10 + 1)/4 = 2.75 ranked value. Round 2.75 to 3 and use the
third ranked value.

To illustrate the computation of the quartiles for the time-to-get-ready data, rank the data from
smallest to largest.
Ranked values:
29

31

35

39

39

40

43

44

44

52

10

Ranks:

The first quartile is the (n + 1)/4 = (10 + 1)/4 = 2.75 ranked value. Using the third rule for quartiles, you round up to the third ranked value. The third ranked value for the get-ready time data
is 35 minutes. You interpret the first quartile of 35 to mean that on 25% of the days the time to
get ready is less than or equal to 35 minutes, and on 75% of the days, the time to get ready is
greater than or equal to 35 minutes.
The third quartile is the 3(n + 1)/4 = 3(10 + 1)/4 = 8.25 ranked value. Using the third rule
for quartiles, you round this down to the eighth ranked value. The eighth ranked value for the
get-ready time data is 44 minutes. You interpret this to mean that on 75% of the days, the time
to get ready is less than or equal to 44 minutes, and on 25% of the days, the time to get ready is
greater than or equal to 44 minutes.

78

CHAPTER THREE Numerical Descriptive Measures

EXAMPLE 3.5

COMPUTING THE QUARTILES

The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the first quartile (Q1) and third quartile (Q3) 2003 return
for the small cap mutual funds with high risk. MUTUALFUNDS2004
SOLUTION Ranked from smallest to largest, the percentage return in 2003 for the nine small
cap mutual funds with high risk is:
Ranked value:
37.3

39.2

44.2

44.5

53.8

56.6

59.3

62.4

66.5

Ranks:

For these data

Q1 =
=

( n + 1)
ranked value
4
9 +1
ranked value = 2.5 ranked value
4

Therefore, using the second rule, Q1 is the 2.5 ranked value, halfway between the second
ranked value and the third ranked value. Since the second ranked value is 39.2 and the third
ranked value is 44.2, the first quartile Q1 is halfway between 39.2 and 44.2. Thus,
Q1 =

39.2 + 44.2
= 41.7
2

To find the third quartile Q3

Q3 =
=

3( n + 1)
ranked value
4
3( 9 + 1)
ranked value = 7.5 ranked value
4

Therefore, using the second rule, Q3 is the 7.5 ranked value, halfway between the seventh
ranked value and the eighth ranked value. Since the seventh ranked value is 59.3 and the eighth
ranked value is 62.4, the third quartile Q3 is halfway between 59.3 and 62.4. Thus,
Q3 =

59.3 + 62.4
= 60.85
2

The first quartile of 41.7 indicates that 25% of the returns in 2003 for small cap high-risk funds
are below or equal to 41.7 and 75% are greater than or equal to 41.7. The third quartile of 60.85
indicates that 75% of the returns in 2003 for small cap high-risk funds are below or equal to
60.85 and 25% are greater than or equal to 60.85.

3.1: Measures of Central Tendency, Variation, and Shape

79

The Geometric Mean

The geometric mean and the geometric rate of return measure the status of an investment over
time. The geometric mean measures the rate of change of a variable over time. Equation (3.5)
defines the geometric mean.
GEOMETRIC MEAN
The geometric mean is the nth root of the product of n values
X G = ( X 1 X 2 L X n )1/ n

(3.5)

Equation (3.6) defines the geometric mean rate of return.

GEOMETRIC MEAN RATE OF RETURN
RG = [(1 + R1 ) (1 + R2 ) L (1 + Rn )]1/ n 1
where

(3.6)

Ri is the rate of return in time period i

To illustrate using these measures, consider an investment of $100,000 that declined to a value
of $50,000 at the end of year 1 and then rebounded back to its original $100,000 value at the
end of year 2. The rate of return for this investment for the two-year period is 0, because the
starting and ending value of the investment is unchanged. However, the arithmetic mean of the
yearly rates of return of this investment is
X =

( 0.50 ) + (1.00 )
= 0.25 or 25%
2

since the rate of return for year 1 is

50, 000 100, 000
R1 =
= 0.50 or 50%

100, 000
and the rate of return for year 2 is
100, 000 50, 000
R2 =
= 1.00 or 100%

50, 000
Using Equation (3.6), the geometric mean rate of return for the two years, is
RG = [(1 + R1 ) (1 + R2 )]1/ n 1
= [(1 + ( 0.50 )) (1 + (1.0 ))]1/ 2 1
= [(0.50 ) ( 2.0 )]1/ 2 1
= [1.0 ]1/ 2 1
= 11 = 0
Thus, the geometric mean rate of return more accurately reflects the (zero) change in the value
of the investment for the two-year period than does the arithmetic mean.

80

CHAPTER THREE Numerical Descriptive Measures

EXAMPLE 3.6

COMPUTING THE GEOMETRIC MEAN RATE OF RETURN

The percentage change in the NASDAQ Composite Index was 31.53% in 2002 and +50.01%
in 2003. Compute the geometric rate of return.
SOLUTION Using Equation (3.6), the geometric mean rate of return in the NASDAQ
Composite Index for the two years is
RG = [(1 + R1 ) (1 + R2 )]1/ n 1
= [(1 + ( 0.3153)) (1 + (0.5001))]1/ 2 1
= [(0.6847) (1.5001)]1/ 2 1
= [1.0271]1/ 2 1
= 1.0135 1 = 0.0135
The geometric rate of return in the NASDAQ Composite Index for the two years is 1.35%.

The Range
The range is the simplest numerical descriptive measure of variation in a set of data.
RANGE
The range is equal to the largest value minus the smallest value.
Range = Xlargest Xsmallest

(3.7)

To determine the range of the times to get ready, you rank the data from smallest to largest:
29 31 35 39 39 40 43 44 44 52
Using Equation (3.7), the range is 52 29 = 23 minutes. The range of 23 minutes indicates that
the largest difference between any two days in the time to get ready in the morning is 23 minutes.

EXAMPLE 3.7

COMPUTING THE RANGE IN THE 2003 RETURN OF SMALL CAP HIGH-RISK

MUTUAL FUNDS
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the range of the 2003 return for the small cap mutual
funds with high risk. MUTUALFUNDS2004
SOLUTION Ranked from the smallest to the largest, the 2003 return for the nine small cap
mutual funds with high risk is:
37.3

39.2

44.2

44.5

53.8

56.6

59.3

62.4

66.5

Therefore, using Equation (3.7) the range = 66.5 37.3 = 29.2.

The largest difference between any two returns for the small cap mutual funds with high
risk is 29.2.

3.1: Measures of Central Tendency, Variation, and Shape

81

The range measures the total spread in the set of data. Although the range is a simple measure of total variation in the data, it does not take into account how the data are distributed
between the smallest and largest values. In other words, the range does not indicate if the values are evenly distributed throughout the data set, clustered near the middle, or clustered near
one or both extremes. Thus, using the range as a measure of variation when at least one value is
an extreme value is misleading.

The Interquartile Range

The interquartile range (also called midspread) is the difference between the third and first
quartiles in a set of data.
INTERQUARTILE RANGE
The interquartile range is the difference between the third quartile and the first quartile.
Interquartile range = Q3 Q1

(3.8)

The interquartile range measures the spread in the middle 50% of the data; therefore, it is not
influenced by extreme values. To determine the interquartile range of the times to get ready
29 31 35 39 39 40 43 44 44 52
you use Equation (3.8) and the earlier results on page 78, Q1 = 35 and Q3 = 44.
Interquartile range = 44 35 = 9 minutes
Therefore, the interquartile range in the time to get ready is 9 minutes. The interval 35 to 44 is
often referred to as the middle fifty.

EXAMPLE 3.8

COMPUTING THE INTERQUARTILE RANGE FOR THE 2003 RETURN

OF SMALL CAP HIGH-RISK MUTUAL FUNDS
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the interquartile range of the 2003 return for the small
cap mutual funds with high risk. MUTUALFUNDS2004
SOLUTION Ranked from smallest to largest, the 2003 return for the nine small cap mutual
funds with high risk is:
37.3

39.2

44.2

44.5

53.8

56.6

59.3

62.4

66.5

Using Equation (3.8) and the earlier results on page 78, Q1 = 41.7 and Q3 = 60.85.
Interquartile range = 60.85 41.7 = 19.15
Therefore, the interquartile range in the 2003 return is 19.15.

Because the interquartile range does not consider any value smaller than Q1 or larger than
Q3, it cannot be affected by extreme values. Summary measures such as the median, Q1, Q3,
and the interquartile range, which cannot be influenced by extreme values, are called resistant
measures.

82

CHAPTER THREE Numerical Descriptive Measures

The Variance and the Standard Deviation

Although the range and the interquartile range are measures of variation, they do not take into
consideration how the values distribute or cluster between the extremes. Two commonly used
measures of variation that take into account how all the values in the data are distributed are the
variance and the standard deviation. These statistics measure the average scatter around
the meanhow larger values fluctuate above it and how smaller values distribute below it.
A simple measure of variation around the mean might take the difference between each
value and the mean and then sum these differences. However, if you did that, you would find
that because the mean is the balance point in a set of data, for every set of data these differences
would sum to zero. One measure of variation that would differ from data set to data set would
square the difference between each value and the mean and then sum these squared differences.
In statistics, this quantity is called a sum of squares (or SS). This sum is then divided by the
number of values minus 1 (for sample data) to get the sample variance (S 2). The square root of
the sample variance is the sample standard deviation (S).
Because the sum of squares are a sum of squared differences that by the rules of arithmetic
will always be nonnegative, neither the variance nor the standard deviation can ever be negative. For most sets of data, the variance and standard deviation will be a positive value, although
both of these statistics will be zero if there is no variation at all in a set of data and each value
in the sample is the same.
For a sample containing n values, X1, X2, X3, . . . , Xn, the sample variance (given by the
symbol S2) is
S2 =

( X1 X )2 + ( X 2 X )2 + L + ( X n X )2
n 1

Equation (3.9) expresses the equation using summation notation.

SAMPLE VARIANCE
The sample variance is the sum of the squared differences around the mean divided by the
sample size minus one.
n

S2 =
where

( X i X )2
i =1

n 1

(3.9)

X = mean
n = sample size
Xi = ith value of the variable X

( X i X )2 = summation of all the squared differences between the Xi values and X

i =1

SAMPLE STANDARD DEVIATION

The sample standard deviation is the square root of the sum of the squared differences
around the mean divided by the sample size minus one.
n

S =

S2 =

( X i X )2
i =1

n 1

(3.10)

3.1: Measures of Central Tendency, Variation, and Shape

83

If the denominator were n instead of n 1, Equation (3.9) [and the inner term in Equation
(3.10)] would calculate the average of the squared differences around the mean. However,
n 1 is used because of certain desirable mathematical properties possessed by the statistic
S2 that make it appropriate for statistical inference (which will be discussed in Chapter 7).
As the sample size increases, the difference between dividing by n or n 1 becomes smaller
and smaller.
You will most likely use the sample standard deviation as your measure of variation
[defined in Equation (3.10)]. Unlike the sample variance, which is a squared quantity, the standard deviation is always a number that is in the same units as the original sample data. The standard deviation helps you to know how a set of data clusters or distributes around its mean. For
almost all sets of data, the majority of the observed values lie within an interval of plus and
minus one standard deviation above and below the mean. Therefore, knowledge of the mean
and the standard deviation usually helps define where at least the majority of the data values
are clustering.
To hand-calculate the sample variance S2 and the sample standard deviation S:
Step 1: Compute the difference between each value and the mean.
Step 2: Square each difference.
Step 3: Add the squared differences.
Step 4: Divide this total by n 1 to get the sample variance.
Step 5: Take the square root of the sample variance to get the sample standard deviation.
Table 3.1 shows the first four steps for calculating the variance and standard deviation
for the getting ready times data with a mean ( X ) equal to 39.6 (see page 74 for the calculation of the mean). The second column of Table 3.1 shows Step 1. The third column of Table
3.1 shows Step 2. The sum of the squared differences (Step 3) is shown at the bottom of
Table 3.1. This total is then divided by 10 1 = 9 to compute the variance (Step 4).

TABLE 3.1
Computing the Variance
of the Getting Ready
Times

X = 39.6
Time
(X)
39
29
43
52
39
44
40
31
44
35

Step 1:
(Xi X )

Step 2:
(Xi X )2

0.60
10.60
3.40
12.40
0.60
4.40
0.40
8.60
4.40
4.60

0.36
112.36
11.56
153.76
0.36
19.36
0.16
73.96
19.36
21.16

Step 3:
Sum:

Step 4:
Divide by (n 1):

412.40

45.82

84

CHAPTER THREE Numerical Descriptive Measures

You can also calculate the variance by substituting values for the terms in Equation (3.9):
n

S2 =

( X i X )2
i =1

n 1

( 39 39.6 ) 2 + ( 29 39.6 ) 2 + L + ( 35 39.6 ) 2

10 1

412.4
9

= 45.82
Because the variance is in squared units (in squared minutes for these data), to compute the
standard deviation you take the square root of the variance. Using Equation (3.10) on page 82,
the sample standard deviation S is
n

S2 =

S =

( X i X )2
i =1

n 1

45.82 = 6.77

This indicates that the get-ready times in this sample are clustering within 6.77 minutes around
the mean of 39.6 minutes (i.e., clustering between X 1S = 32.83 and X + 1S = 46.37). In
fact, 7 out of 10 get-ready times lie within this interval.
Using the second column of Table 3.1, you can also calculate the sum of the differences
between each value and the mean to be zero. For any set of data, this sum will always be zero:
n

( X i X ) = 0 for all sets of data

i =1

This property is one of the reasons that the mean is used as the most common measure of central tendency.

EXAMPLE 3.9

COMPUTING THE VARIANCE AND STANDARD DEVIATION OF THE 2003 RETURN

OF SMALL CAP HIGH-RISK MUTUAL FUNDS
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the variance and standard deviation of the 2003 return
for the small cap mutual funds with high risk. MUTUALFUNDS2004
SOLUTION Table 3.2 illustrates the computation of the variance and standard deviation for
the return in 2003 for the small cap mutual funds with high risk. Using Equation (3.9) on page 82
n

S2 =

( X i X )2
i =1

n 1

( 44.5 51.53) 2 + ( 39.2 51.53) 2 + L + ( 66.5 51.53) 2

9 1

891.16
8

= 111.395

3.1: Measures of Central Tendency, Variation, and Shape

TABLE 3.2
Computing the Variance
of the 2003 Return for
the Small Cap Mutual
Funds with High Risk

85

X = 51.5333
Return
2003

Step 1:
(Xi X )

Step 2:
(Xi X )2

44.5
39.2
62.4
59.3
56.6
53.8
37.3
44.2
66.5

7.0333
12.3333
10.8667
7.7667
5.0667
2.2667
14.2333
7.3333
14.9667

49.4678
152.1111
118.0844
60.3211
25.6711
5.1378
202.5878
53.7778
224.0011

Step 3:
Sum:

Step 4:
Divide by (n 1):

891.16

111.395

Using Equation (3.10) on page 82, the sample standard deviation S is

n

S =

S2 =

( X i X )2
i =1

n 1

= 111.395 = 10.55

The standard deviation of 10.55 indicates that the 2003 returns for the small cap mutual
funds with high risk are clustering within 10.55 around the mean of 51.53 (i.e., clustering
between X 1S = 40.98 and X + 1S = 62.08). In fact, 55.6% (5 out of 9) of the 2003 returns
lie within this interval.

The following summarizes the characteristics of the range, interquartile range, variance,
and standard deviation.

The more spread out, or dispersed, the data are, the larger the range, interquartile range,
variance, and standard deviation.
The more concentrated, or homogeneous the data are, the smaller the range, interquartile
range, variance, and standard deviation.
If the values are all the same (so that there is no variation in the data), the range, interquartile range, variance, and standard deviation will all equal zero.
None of the measures of variation (the range, interquartile range, standard deviation, and
variance) can ever be negative.

The Coefficient of Variation

Unlike the previous measures of variation presented, the coefficient of variation is a relative
measure of variation that is always expressed as a percentage rather than in terms of the units of
the particular data. The coefficient of variation, denoted by the symbol CV, measures the scatter in the data relative to the mean.

86

CHAPTER THREE Numerical Descriptive Measures

COEFFICIENT OF VARIATION
The coefficient of variation is equal to the standard deviation divided by the mean,
multiplied by 100%.
S
CV = 100%
X
where

(3.11)

S = sample standard deviation

X = sample mean

For the sample of 10 get-ready times, since X = 39.6 and S = 6.77, the coefficient of variation is
S
6.77
CV = 100% =
100% = 17.10%
X
39.6
For the get-ready times, the standard deviation is 17.1% of the size of the mean.
You will find the coefficient of variation very useful when comparing two or more sets of
data that are measured in different units as Example 3.10 illustrates.

EXAMPLE 3.10

COMPARING TWO COEFFICIENTS OF VARIATION WHEN TWO VARIABLES

HAVE DIFFERENT UNITS OF MEASUREMENT
The operations manager of a package delivery service is deciding on whether to purchase a new
fleet of trucks. When packages are stored in the trucks in preparation for delivery, you need to
consider two major constraintsthe weight (in pounds) and the volume (in cubic feet) for each
item.
The operations manager samples 200 packages, and finds that the mean weight is 26.0
pounds, with a standard deviation of 3.9 pounds, and the mean volume is 8.8 cubic feet, with a
standard deviation of 2.2 cubic feet. How can the operations manager compare the variation of
the weight and the volume?
SOLUTION Because the measurement units differ for the weight and volume constraints, the
operations manager should compare the relative variability in the two types of measurements.
For weight, the coefficient of variation is
3.9
CVW =
100% = 15%
26.0
For volume, the coefficient of variation is
2.2
CVV =
100% = 25.0%
8.8
Thus, relative to the mean, the package volume is much more variable than the package weight.

Z Scores
An extreme value or outlier is a value located far away from the mean. Z scores are useful in
identifying outliers. The larger the Z score, the farther the distance from the value to the mean.
The Z score is the difference between the value and the mean, divided by the standard deviation.

3.1: Measures of Central Tendency, Variation, and Shape

87

Z SCORES
Z =

X X
S

(3.12)

For the time to get ready in the morning data, the mean is 39.6 minutes and the standard deviation is 6.77 minutes. The time to get ready on the first day is 39.0 minutes. You compute the Z
score for day 1 from
Z =
=

X X
S
39.0 39.6
6.77

= 0.09
Table 3.3 shows the Z scores for all 10 days. The largest Z score is 1.83 for day 4 on which
the time to get ready was 52 minutes. The lowest Z score was 1.57 for day 2 on which the
time to get ready was 29 minutes. As a general rule, a Z score is considered an outlier if it
is less than 3.0 or greater than +3.0. None of the times met that criterion to be considered
outliers.

TABLE 3.3
Z Scores for the
10 Get-Ready Times

Mean
Standard deviation

EXAMPLE 3.11

Time (X)

Z Score

39
29
43
52
39
44
40
31
44
35
39.6
6.77

0.09
1.57
0.50
1.83
0.09
0.65
0.06
1.27
0.65
0.68

COMPUTING THE Z SCORES OF THE 2003 RETURN OF SMALL CAP HIGH-RISK

MUTUAL FUNDS
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the Z scores of the 2003 return for the small cap mutual
funds with high risk. MUTUALFUNDS2004
SOLUTION Table 3.4 illustrates the Z scores of the 2003 return for the small cap mutual
funds with high risk. The largest Z score is 1.42 for a percentage return of 66.5. The lowest Z
score is 1.35 for a percentage return of 37.3. As a general rule, a Z score is considered an outlier if it is less than -3.0 or greater than +3.0. None of the percentage returns met that criterion
to be considered outliers.

88

CHAPTER THREE Numerical Descriptive Measures

TABLE 3.4
Z Scores of the 2003
Return for the Small
Cap Mutual Funds
with High Risk

Mean
Standard Deviation

Return 2003

Z Scores

44.5
39.2
62.4
59.3
56.6
53.8
37.3
44.2
66.5
51.53
10.55

0.67
1.17
1.03
0.74
0.48
0.21
1.35
0.69
1.42

Shape
A third important property that describes a set of numerical data is shape. Shape is the pattern
of the distribution of data values throughout the entire range of all the values. A distribution
will either be symmetrical, when low and high values balance each other out, or skewed, not
symmetrical and showing an imbalance of low values or high values.
Shape influences the relationship of the mean to the median in the following ways:

Mean < median; negative or left-skewed

Mean = median; symmetric or zero skewness
Mean > median; positive or right-skewed
Figure 3.1 depicts three data sets, each with a different shape.

FIGURE 3.1
A Comparison of Three
Data Sets Differing
in Shape
Panel A
Negative, or left-skewed

Panel B
Symmetrical

Panel C
Positive, or right-skewed

The data in panel A are negative, or left-skewed. In this panel, most of the values are in the
upper portion of the distribution. There is a long tail and distortion to the left that is caused by
some extremely small values. These extremely small values pull the mean downward so that the
mean is less than the median.
The data in panel B are symmetrical. Each half of the curve is a mirror image of the other
half of the curve. The low and high values on the scale balance, and the mean equals the median.
The data in panel C are positive, or right-skewed. In this panel, most of the values are in
the lower portion of the distribution. There is a long tail on the right of the distribution and a
distortion to the right that is caused by some extremely large values. These extremely large values pull the mean upward so that the mean is greater than the median.

Visual Explorations: Exploring Descriptive Statistics

Use the Visual Explorations Descriptive Statistics procedure to see the effect of changing data
values on measures of central tendency, variation, and shape. Open the Visual Explorations.xla
macro workbook and select VisualExplorations  Descriptive Statistics from the Microsoft
Excel menu bar. Read the instructions in the popup box (see illustration on page 89) and click OK
to examine a dot scale diagram for the sample of 10 get-ready times used throughout this chapter.

3.1: Measures of Central Tendency, Variation, and Shape

89

Experiment by entering an extreme value such as 10 minutes into one of the tinted cells of
column A. Which measures are affected by this change? Which ones are not? You can flip
between the before and above diagrams by repeatedly pressing Crtl-Z (undo) followed by
Crtl-Y (redo) to help see the changes the extreme value caused in the diagram.

Microsoft Excel Descriptive Statistics Output

The Microsoft Excel Data Analysis ToolPak generates the mean, median, mode, standard deviation, variance, range, minimum, maximum, and count (sample size) on a single worksheet, all
of which have been discussed in this section. In addition, Excel computes the standard error,
along with statistics for kurtosis and skewness. The standard error is the standard deviation
divided by the square root of the sample size and will be discussed in Chapter 7. Skewness
measures the lack of symmetry in the data and is based on a statistic that is a function of the
cubed differences around the mean. A skewness value of zero indicates a symmetric distribution. Kurtosis measures the relative concentration of values in the center of the distribution as
compared with the tails and is based on the differences around the mean raised to the fourth
power. This measure is not discussed in this text (see reference 2).
From Figure 3.2, the Excel descriptive statistics output for the 2003 return of the funds
based on risk level, there appears to be slight differences in the 2003 percentage return for the
FIGURE 3.2
Microsoft Excel
Descriptive Statistics
of the 2003 Returns
Based on Risk Level

90

CHAPTER THREE Numerical Descriptive Measures

three risk levels. High-risk funds had a slightly higher mean and median than did low-risk and
average-risk funds. There was very little difference in the standard deviations of the three
groups.

Minitab Descriptive Statistics Output

For descriptive statistics, Minitab computes the sample size (labeled as N), the mean, median,
standard deviation (labeled StDev), minimum, maximum, coefficient of variation (labeled
CoefVar), first and third quartiles, range, and interquartile range (labeled IQR), all of which
have been discussed in this section.
From Figure 3.3, the Minitab descriptive statistics output for the 2003 return of the funds
based on risk level, there appears to be slight differences in the 2003 percentage return for the
three risk levels. High-risk funds had a slightly higher mean, median, and quartiles than did
low-risk and average-risk funds. There was very little difference in the standard deviations or
interquartile ranges of the three groups.
FIGURE 3.3
Minitab Descriptive
Statistics of the 2003
Returns Based
on Risk Level

PROBLEMS FOR SECTION 3.1

Learning the Basics
PH Grade
ASSIST

3.1 The following is a set of data from a sample

of n = 5:
7 4 9 8 2

a. Compute the mean, median, and mode.

b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Compute the Z scores. Are there any outliers?
d. Describe the shape of the data set.
PH Grade
ASSIST

3.2 The following is a set of data from a sample

of n = 6:
7 4 9 7 3 12

a. Compute the mean, median, and mode.

b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.

c. Compute the Z scores. Are there any outliers?

d. Describe the shape of the data set.
3.3 The following set of data is from a sample
of n = 7:
12 7 4 9 0 7 3
a. Compute the mean, median, and mode.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Describe the shape of the data set.
PH Grade
ASSIST

3.4 The following is a set of data from a sample

of n = 5:
7 5 8 7 9
a. Compute the mean, median, and mode.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Describe the shape of the data set.
PH Grade
ASSIST

3.1: Measures of Central Tendency, Variation, and Shape

3.5 Suppose that the rate of return for a particular stock during the past two years was 10% and
30%. Compute the geometric mean rate of return.
(Note: A rate of return of 10% is recorded as 0.10 and a rate
of return of 30% is recorded as 0.30.)
PH Grade
ASSIST

Applying the Concepts

Problems 3.63.20 can be solved manually or by using
Microsoft Excel, Minitab, or SPSS.

3.6 The operations manager of a plant that manufactures tires wants to compare the actual inner
diameter of two grades of tires, each of which is
expected to be 575 millimeters. A sample of five tires of
each grade was selected, and the results representing the
inner diameters of the tires, ranked from smallest to largest,
are as follows:
PH Grade
ASSIST

Grade X

Grade Y

568 570 575 578 584 573 574 575 577 578
a. For each of the two grades of tires, compute the mean,
median, and standard deviation.
b. Which grade of tire is providing better quality? Explain.
c. What would be the effect on your answers in (a) and (b) if
the last value for grade Y were 588 instead of 578? Explain.
SELF
Test

3.7 The following data represent the total fat for

burgers and chicken items from a sample of fastfood chains. FASTFOOD
Burgers
19 31 34 35 39 39 43
Chicken

7 9 15 16 16 18 22 25 27 33 39
Source: Extracted from Quick Bites, Copyright 2001 by
Consumers Union of U.S., Inc., Yonkers, NY 107031057. Adapted
with permission from Consumer Reports, March 2001, 46.

For the burgers and chicken items separately:

a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the variance, standard deviation, range,
interquartile range, and coefficient of variation.
c. Are the data skewed? If so, how?
d. Based on the results of (a) through (c), what conclusions
can you reach concerning the differences in total fat of
burgers and chicken items?
3.8 The median price of a home in December 2003 rose to
$173,200, an increase of 6.7% from December 2002. For
the full year, sales hit a record 6.1 million homes (James R.
Hagerty, Housing Prices Continue to Rise, The Wall
Street Journal, January 27, 2004, D1).
a. Describe the shape of the distribution of the price of
homes sold.
b. Why do you think the article reports the median home
price and not the mean home price?

91

3.9 In the 20022003 academic year, many public universities in the United States raised tuition and fees due to a
decrease in state subsidies (Mary Beth Marklein, Public
Universities Raise Tuition, Feesand Ire, USA Today,
August 8, 2002, 1A2A). The following represents the
change in the cost of tuition, a shared dormitory room, and
the most popular meal plan between the 20012002 academic year and the 20022003 academic year for a sample of
10 public universities. COLLEGECOST
University

Change in Cost ($)

University of California, Berkeley

University of Georgia, Athens
University of Illinois, UrbanaChampaign
Kansas State University, Manhattan
University of Maine, Orono
University of Mississippi, Oxford
University of New Hampshire, Durham
Ohio State University, Columbus
University of South Carolina, Columbia
Utah State University, Logan

1,589
593
1,223
869
423
1,720
708
1,425
922
308

a. Compute the mean, median, first quartile, and third

quartile.
b. Compute the variance, standard deviation, range, interquartile range, coefficient of variation, and Z scores.
c. Are the data skewed? If so, how?
d. Based on the results of (a) through (c), what conclusions
can you reach concerning the change in costs between
the 20012002 and 20022003 academic years?
3.10 The following data COFFEEDRINK represent the calories and fat (in grams) of 16-ounce iced coffee drinks at
Dunkin Donuts and Starbucks.
Product
Dunkin Donuts Iced Mocha Swirl latte
(whole milk)
Starbucks Coffee Frappuccino blended
coffee
Dunkin Donuts Coffee Coolatta (cream)
Starbucks Iced Coffee Mocha Expresso
(whole milk and whipped cream)
Starbucks Mocha Frappuccino blended
coffee (whipped cream)
Starbucks Chocolate Brownie
Frappuccino blended coffee
(whipped cream)
Starbucks Chocolate Frappuccino
Blended Crme (whipped cream)

Calories Fat
240

8.0

260
350

3.5
22.0

350

20.0

420

16.0

510

22.0

530

19.0

Source: Extracted from Coffee as Candy at Dunkin Donuts and

Starbucks, Copyright 2004 by Consumers Union of U.S., Inc.,
Yonkers, NY 107031057. Adapted with permission from Consumer
Reports, June 2004, 9.

92

CHAPTER THREE Numerical Descriptive Measures

For each variable (calories and fat),

a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the variance, standard deviation, range,
interquartile range, coefficient of variation, and Z
scores. Are there any outliers? Explain.
c. Are the data skewed? If so, how?
d. Based on the results of (a) through (c), what conclusions
can you reach concerning the calories and fat in iced
coffee drinks at Dunkin Donuts and Starbucks?
3.11 The following data represent the daily hotel cost and
rental car cost for 20 U. S. cities during a week in October
2003. HOTEL-CAR
City
San Francisco
Los Angeles
Seattle
Phoenix
Denver
Dallas
Houston
Minneapolis
Chicago
St. Louis
New Orleans
Detroit
Cleveland
Atlanta
Orlando
Miami
Pittsburgh
Boston
New York
Washington, D.C.

Hotel

Cars

205
179
185
210
128
145
177
117
221
159
205
128
165
180
198
158
132
283
269
204

47
41
49
38
32
48
49
41
56
41
50
32
34
46
41
40
39
67
69
40

Source: Extracted from The Wall Street Journal, October 10, 2003, W4.

For each variable (hotel cost and rental car cost),

a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the variance, standard deviation, range,
interquartile range, coefficient of variation, and Z
scores. Are there any outliers? Explain.
c. Are the data skewed? If so, how?
d. Based on the results of (a) through (c), what conclusions
can you reach concerning the daily cost of a hotel and
rental car?
3.12 The cost of 14 models of 3-megapixel digital cameras at a camera specialty store during 2003 was as follows.
CAMERA

340 450 450 280 220 340 290

370 400 310 340 430 270 380

a. Compute the mean, median, first quartile, and third

quartile.
b. Compute the variance, standard deviation, range,
interquartile range, coefficient of variation, and Z
scores. Are there any outliers? Explain.
c. Are the data skewed? If so, how?
d. Based on the results of (a) through (c), what conclusions
can you reach concerning the price of 3-megapixel digital cameras at a camera specialty store during 2003?
3.13 A software development and consulting firm located
in the Phoenix metropolitan area develops software for
supply chain management systems using systematic software reuse. Instead of starting from scratch when writing
and developing new custom software systems, the firm
uses a database of reusable components totaling more than
2,000,000 lines of code collected from 10 years of continuous reuse effort. Eight analysts at the firm were asked to
estimate the reuse rate when developing a new software
system. The following data are given as a percentage of the
total code written for a software system that is part of the
reuse database. REUSE
50 62.5 37.5 75.0 45.0 47.5 15.0 25.0
Source: M. A. Rothenberger, and K. J. Dooley, A Performance
Measure for Software Reuse Projects, Decision Sciences, 30(Fall
1999), 11311153.

a. Compute the mean, median, and mode.

b. Compute the range, variance, and standard deviation.
c. Interpret the summary measures calculated in (a) and (b).
3.14 A manufacturer of flashlight batteries took a sample
of 13 batteries from a days production and used them continuously until they were drained. The numbers of hours
they were used until failure were: BATTERIES
342 426 317 545 264 451
1,049 631 512 266 492 562 298
a. Compute the mean, median, and mode. Looking at the
distribution of times to failure, which measures of location do you think are most appropriate and which least
appropriate to use for these data? Why?
b. Calculate the range, variance, and standard deviation.
c. What would you advise if the manufacturer wanted to be
able to say in advertisements that these batteries should
last 400 hours? (Note: There is no right answer to this
question; the point is to consider how to make such a
statement precise.)
d. Suppose that the first value was 1,342 instead of 342.
Repeat (a) through (c), using this value. Comment on
the difference in the results.
3.15 A bank branch located in a commercial district of a city
has developed an improved process for serving customers
during the noon to 1:00 P.M. lunch period. The waiting time in
minutes (defined as the time the customer enters the line to

93

3.1: Measures of Central Tendency, Variation, and Shape

when he or she reaches the teller window) of all customers

during this hour is recorded over a period of one week. A random sample of 15 customers is selected, and the results are as
follows: BANK1
4.21 5.55 3.02 5.13 4.77 2.34 3.54
3.20 4.50 6.10 0.38 5.12 6.46 6.19 3.79
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the variance, standard deviation, range,
interquartile range, coefficient of variation, and Z
scores. Are there any outliers? Explain.
c. Are the data skewed? If so, how?
d. As a customer walks into the branch office during the
lunch hour, she asks the branch manager how long she can
expect to wait. The branch manager replies, Almost certainly less than five minutes. On the basis of the results of
(a) and (b), evaluate the accuracy of this statement.
3.16 Suppose that another branch, located in a residential
area, is also concerned with the noon to 1 P.M. lunch hour.
The waiting time in minutes (defined as the time the customer enters the line to the time he or she reaches the teller
window) of all customers during these hours is recorded
over a period of one week. A random sample of 15 customers is selected, and the results are as follows: BANK2
9.66 5.90 8.02 5.79 8.73 3.82 8.01
8.35 10.49 6.68 5.64 4.08 6.17 9.91 5.47
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the variance, standard deviation, range,
interquartile range, and coefficient of variation. Are
there any outliers? Explain.
c. Are the data skewed? If so, how?
d. As a customer walks into the branch office during the
lunch hour, he asks the branch manager how long he can
expect to wait. The branch manager replies, Almost certainly less than five minutes. On the basis of the results
of (a) and (b), evaluate the accuracy of this statement.
3.17 China is the fastest-growing market for
passenger car sales and fourth biggest after the
United States, Japan, and Germany. Passenger
car sales increased 61% in 2002 and 55% in 2003 (Peter
Wonacott, A Fear Amid Chinas Car Boom, The Wall
Street Journal, February 2, 2004, A17). Compute the geometric mean rate of increase. (Hint: Denote an increase of
61% as R1 = 0.61.)
SELF
Test

3.18 The time period from 2000 to 2003 saw a great deal
of volatility in the value of stocks. The data in the following
table STOCKRETURN represent the total rate of return of the
Dow Jones Industrial Index, the Standard & Poors 500, the
Russell 2000 Index, and the Wilshire 5000 Index from
2000 to 2003.

Year

DJIA

SP500

Russell2000

Wilshire5000

2003
2002
2001
2000

25.30
15.01
5.44
6.20

26.40
22.10
11.90
9.10

45.40
21.58
1.03
3.02

29.40
20.90
10.97
10.89

Source: Extracted from The Wall Street Journal, January 2, 2004.

a. Calculate the geometric rate of return for the Dow Jones

Industrial Index, the Standard & Poors 500, the Russell
2000 Index, and the Wilshire 5000 Index.
b. What conclusions can you reach concerning the geometric rates of return of the four stock indexes?
c. Compare the results of (b) to those of problems 3.19 (b)
and 3.20 (b).
3.19 The time period from 2000 to 2003 saw a great deal of
volatility in the value of investments. The data in the following table BANKRETURN represent the total rate of return of the
one-year certificate of deposit, the 30-month certificate of
deposit, and the money market deposit from 2000 to 2003.
Year

One Year

30 Month

Money Market

2003
2002
2001
2000

1.20
1.98
3.60
5.46

1.76
2.74
3.97
5.64

0.61
1.02
1.73
2.09

Source: Extracted from The Wall Street Journal, January 2, 2004.

a. Calculate the geometric rate of return for the one-year

certificate of deposit, the 30-month certificate of
deposit, and the money market deposit.
b. What conclusions can you reach concerning the geometric rates of return of the three deposits?
c. Compare the results of (b) to those of problems 3.18 (b)
and 3.20 (b).
3.20 The time period from 2000 to 2003 saw a great deal
of volatility in the value of metals. The data in the following table METALRETURN represent the total rate of return for
platinum, gold, and silver from 2000 to 2003.
Year

Platinum

Gold

Silver

2003
2002
2001
2000

34.2
24.5
21.3
23.3

19.5
24.5
1.2
1.8

24.0
5.5
3.0
5.9

Source: Extracted from The Wall Street Journal, January 2, 2004.

a. Calculate the geometric rate of return for platinum,

gold, and silver.
b. What conclusions can you reach concerning the geometric rates of return of the three metals?
c. Compare the results of (b) to those of problems 3.18 (b)
and 3.19 (b).

94

CHAPTER THREE Numerical Descriptive Measures

3.2

NUMERICAL DESCRIPTIVE MEASURES FOR A POPULATION

Section 3.1 presented various statistics that described the properties of central tendency, variation, and shape for a sample. If your data set represents numerical measurements for an entire
population, you need to calculate and interpret parameters, summary measures for a population. In this section, you will learn about three descriptive population parameters, the population mean, population variance, and population standard deviation.
To help illustrate these parameters, first review Table 3.5 that contains the five biggest
bond funds (in terms of total assets) as of March 1, 2004. The 52-week return for each of these
funds is also listed. LARGEST BONDS.

Bond Fund

TABLE 3.5
2003 Return for the
Population Consisting
of the Five Largest
Bond Funds

52-Week Return (in %)

Vanguard GNMA
Vanguard Total Bond Index
Pimco Total Return Admin
Pimco Total Return Instl
America Bond Fund

3.8
6.5
7.0
7.3
12.9

Source: Extracted from The Wall Street Journal, March 25, 2004, C2.

The Population Mean

The population mean is represented by the symbol , the Greek lowercase letter mu. Equation
(3.13) defines the population mean.

POPULATION MEAN
The population mean is the sum of the values in the population divided by the population
size N.
N

=
where

Xi
i =1

(3.13)

= population mean
Xi = ith value of the variable X

X i = summation of all Xi values in the population

i =1

To compute the mean return for the population of bond funds given in Table 3.5, use
Equation (3.13),
N

Xi
i =1

3.8 + 6.5 + 7.0 + 7.3 + 12.9

37.5
=
= 7.5
5
5

Thus, the mean 2003 return for these bond funds is 7.5%.

3.2: Numerical Descriptive Measures for a Population

95

The Population Variance and Standard Deviation

The population variance and the population standard deviation measure variation in a
population. Like the related sample statistics, the population standard deviation is the square
root of the population variance. The symbol 2, the Greek lowercase letter sigma squared, represents the population variance and the symbol , the Greek lowercase letter sigma, represents
the population standard deviation. Equations (3.14) and (3.15) define these parameters. The
denominators for the right-side terms in these equations use N and not the (n 1) term that is
used in the equations for the sample variance and standard deviation [see Equations (3.9) and
(3.10) on page 82].

POPULATION VARIANCE
The population variance is the sum of the squared differences around the population mean
divided by the population size N.
N

2 =
where

( X i )2
i =1

(3.14)

= population mean
Xi = ith value of the variable X

( X i )2 = summation of all the squared differences between the Xi values and

i =1

POPULATION STANDARD DEVIATION

N

( X i )2
i =1

(3.15)

To compute the population variance for the data of Table 3.5 on page 94, you use Equation
(3.14),

2 =

( X i )2
i =1

( 3.8 7.5) 2 + (6.5 7.5) 2 + ( 7.0 7.5) 2 + ( 7.3 7.5) 2 + (12.9 7.5) 2
5

13.69 + 1.00 + 0.25 + 0.04 + 29.16

5

44.14
= 8.828
5

96

CHAPTER THREE Numerical Descriptive Measures

Thus, the variance of the returns is 8.828 squared percentage return. The squared units
make the variance hard to interpret. You should use the standard deviation that uses the original
units of the data (percentage return). From Equation (3.15),
N

2 =

( X i )2
i =1

8.828 = 2.97

Therefore, the typical 2003 return differs from the mean of 7.5 by approximately 2.97. This
large amount of variation suggests that these large bond funds produce results that differ greatly.

The Empirical Rule

In most data sets, a large portion of the values tend to cluster somewhat near the median. In
right-skewed data sets, this clustering occurs to the left of the mean, that is, at a value less than
the mean. In left-skewed data sets, the values tend to cluster to the right of the mean, that is,
at a value greater than the mean. In symmetrical data sets, where the median and mean are
the same, the values often tend to cluster around the median and mean producing a bellshaped distribution. You can use the empirical rule to examine the variability in bell-shaped
distributions:

Approximately 68% of the values are within a distance of 1 standard deviation from the
mean.
Approximately 95% of the values are within a distance of 2 standard deviations from the
mean.
Approximately 99.7% are within a distance of 3 standard deviations from the mean.

The empirical rule helps you measure how the values distribute above and below the mean.
This can help you to identify outliers when analyzing a set of numerical data. The empirical rule
implies that for bell-shaped distributions only about one out of 20 values will be beyond two
standard deviations from the mean in either direction. As a general rule, you can consider values not found in the interval 2 as potential outliers. The rule also implies that only about
three in 1,000 will be beyond three standard deviations from the mean. Therefore, values not
found in the interval 3 are almost always considered outliers. For heavily skewed data sets,
or those not appearing bell-shaped for any other reason, the Chebyshev rule discussed on page
97 should be applied instead of the empirical rule.

EXAMPLE 3.12

USING THE EMPIRICAL RULE

A population of 12-ounce cans of cola is known to have a mean fill-weight of 12.06 ounces and
a standard deviation of 0.02. The population is also known to be bell-shaped. Describe the distribution of fill-weights. Is it very likely that a can will contain less than 12 ounces of cola?
SOLUTION

= 12.06 0.02 = (12.04, 12.08 )

2 = 12.06 2(0.02 ) = (12.02, 12.10 )
3 = 12.06 3( 0.02 ) = (12.00, 12.12 )

Using the empirical rule, approximately 68% of the cans will contain between 12.04 and 12.08
ounces, approximately 95% will contain between 12.02 and 12.10 ounces, and approximately
99.7% will contain between 12.00 and 12.12 ounces. Therefore, it is highly unlikely that a can
will contain less than 12 ounces.

3.2: Numerical Descriptive Measures for a Population

97

The Chebyshev Rule

The Chebyshev rule (reference 1) states that for any data set, regardless of shape, the percentage of values that are found within distances of k standard deviations from the mean must be
at least
(1 1/k2) 100%
You can use this rule for any value of k greater than 1. Consider k = 2. The Chebyshev rule
states that at least [1 (1/2)2] 100% = 75% of the values must be found within 2 standard
deviations of the mean.
The Chebyshev rule is very general and applies to any type of distribution. The rule indicates at least what percentage of the values fall within a given distance from the mean.
However, if the data set is approximately bell-shaped, the empirical rule will more accurately
reflect the greater concentration of data close to the mean. Table 3.6 compares the Chebyshev
and empirical rules.

TABLE 3.6
How Data Vary Around
the Mean

% of Values Found in Intervals Around the Mean

Interval
( , + )
( 2, + 2)
( 3, + 3)

EXAMPLE 3.13

Chebyshev
(for any distribution)

Empirical Rule
(bell-shaped distribution)

At least 0%
At least 75%
At least 88.89%

Approximately 68%
Approximately 95%
Approximately 99.7%

USING THE CHEBYSHEV RULE

As in Example 3.12, a population of 12-ounce cans of cola is known to have a mean fill-weight
of 12.06 ounces and a standard deviation of 0.02. However, the shape of the population is
unknown and you cannot assume that it is bell-shaped. Describe the distribution of fill-weights.
Is it very likely that a can will contain less than 12 ounces of cola?
SOLUTION

= 12.06 0.02 = (12.04, 12.08 )

2 = 12.06 2( 0.02 ) = (12.02, 12.10 )
3 = 12.06 3( 0.02 ) = (12.00, 12.12 )

Because the distribution may be skewed, you cannot use the empirical rule. Using the
Chebyshev rule, you cannot say anything about the percentage of cans containing between
12.04 and 12.08 ounces. You can state that at least 75% of the cans will contain between 12.02
and 12.10 ounces, and at least 88.89% will contain between 12.00 and 12.12 ounces. Therefore,
between 0 and 11.11% of the cans contain less than 12 ounces.

You can use these two rules for understanding how data are distributed around the mean
when you have sample data. In each case, use the value you calculated for X in place of
and the value you calculated for S in place of . The results you compute using the sample
statistics are approximations since you used sample statistics ( X , S) and not population parameters (, ).

98

CHAPTER THREE Numerical Descriptive Measures

PROBLEMS FOR SECTION 3.2

Learning the Basics
3.21 The following is a set of data for a population with N = 10:
7 5 11 8 3 6 2 1 9 8
a. Compute the population mean.
b. Compute the population standard deviation.
PH Grade
ASSIST

3.22 The following is a set of data for a population with N = 10:

7 5 6 6 6 4 8 6 9 3
a. Compute the population mean.
b. Compute the population standard deviation.
PH Grade
ASSIST

Applying the Concepts

3.23 The following data represent the quarterly
sales tax receipts (in thousands of dollars) submitted to the comptroller of the Village of Fair
Lake for the period ending March 2004 by all 50 business
establishments in that locale: TAX
SELF
Test

10.3
13.0
13.0
8.0
11.1
11.6
10.0
12.5
9.3
10.5

11.1
6.7
11.2
11.8
10.2
15.1
12.9
9.3
11.5
7.6

9.6
11.0
7.3
8.7
11.1
12.5
9.2
10.4
10.7
10.1

9.0
8.4
5.3
10.6
9.9
6.5
10.0
12.7
11.6
8.9

14.5
10.3
12.5
9.5
9.8
7.5
12.8
10.5
7.8
8.6

a. Compute the mean, variance, and standard deviation for

this population.
b. What proportion of these businesses have quarterly
sales tax receipts within 1, 2, or 3 standard deviations of the mean?
c. Compare and contrast your findings with what would be
expected on the basis of the empirical rule. Are you surprised at the results in (b)?
3.24 Consider a population of 1,024 mutual
funds that primarily invested in large companies.
You determined that , the mean one-year total
percentage return achieved by all the funds, is 8.20 and
that , the standard deviation, is 2.75. In addition, suppose
you determined that the range in the one-year total returns
is from 2.0 to 17.1 and that the quartiles are, respectively, 5.5 (Q1) and 10.5 (Q3). According to the empirical
rule, what percentage of these funds is expected to be
a. within 1 standard deviation of the mean?
PH Grade
ASSIST

b. within 2 standard deviations of the mean?

c. According to the Chebyshev rule, what percentage of
these funds are expected to be within 1, 2, or 3 standard deviations of the mean?
d. According to the Chebyshev rule, at least 93.75% of
these funds are expected to have one-year total returns
between what two amounts?
3.25 The following table ASSETS represents the assets in
billions of dollars of the five largest bond funds.
Bond Fund
Vanguard GNMA
Vanguard Total Bond Mkt. Index
Bond Fund of America A
Franklin Calif. Tax-Free Inc. A
Vanguard Short-Term Corp.

Assets (Billions $)
19.5
16.8
13.7
12.8
10.9

a. Compute the mean for this population of the five largest

bond funds. Interpret this parameter.
b. Compute the variance and standard deviation for this
population. Interpret these parameters.
c. Is there a lot of variability in the assets of the bond funds?
3.26 The data in the file ENERGY contains the per capita
energy consumption in kilowatt hours for each of the 50
states and the District of Columbia during 1999.
a. Compute the mean, variance, and standard deviation for
the population.
b. What proportion of these states has average per capita
energy consumption within 1 standard deviation of the
mean, within 2 standard deviations of the mean, and
within 3 standard deviations of the mean?
c. Compare and contrast your findings versus what would
be expected based on the empirical rule. Are you surprised at the results in (b)?
d. Do (a) through (c) with the District of Columbia
removed. How have the results changed?
3.27 The data in the file DOWRETURN give the 10-year
annualized return (19942003) for the 30 companies in the
Dow Jones Industrials.
a. Compute the mean for this population. Interpret this
number.
b. Compute the variance and standard deviation for this
population. Interpret the standard deviation.
c. Use the empirical rule or the Chebyshev rule, whichever is
appropriate, to further explain the variation in this data set.
d. Using the results in (c), are there any outliers? Explain.

3.3: Computing Numerical Descriptive Measures from a Frequency Distribution

3.3

99

COMPUTING NUMERICAL DESCRIPTIVE MEASURES

FROM A FREQUENCY DISTRIBUTION
Sometimes you have only a frequency distribution, not the raw data. When this occurs, you can
compute approximations to the mean and the standard deviation.
When you have data from a sample that has been summarized into a frequency distribution, you can compute an approximation of the mean by assuming that all values within each
class interval are located at the midpoint of the class.
APPROXIMATING THE MEAN FROM A FREQUENCY DISTRIBUTION
c

X =
where

mj f j
j =1

(3.16)

X = sample mean
n = number of values or sample size
c = number of classes in the frequency distribution
mj = midpoint of the jth class
fj = numbers of values in the jth class

To calculate the standard deviation from a frequency distribution, you assume that all values
within each class interval are located at the midpoint of the class.
APPROXIMATING THE STANDARD DEVIATION FROM A FREQUENCY DISTRIBUTION
c

S =

( m j X )2 f j
j =1

(3.17)

n 1

Example 3.14 illustrates the computation of the mean and the standard deviation from a
frequency distribution.

EXAMPLE 3.14

APPROXIMATING THE MEAN AND STANDARD DEVIATION

FROM A FREQUENCY DISTRIBUTION
Consider the frequency distribution of the 2003 return of growth funds (Table 3.7). Compute
the mean and standard deviation.

TABLE 3.7
Frequency Distribution
of the 2003 Return for
Growth Mutual Funds

Annual Percentage 2003 Return

10 but less than 20
20 but less than 30
30 but less than 40
40 but less than 50
50 but less than 60
60 but less than 70
Total

Frequency
2
9
13
15
5
5
49

100

CHAPTER THREE Numerical Descriptive Measures

SOLUTION The computations that you need to calculate the approximations of the mean and
standard deviation of the 2003 return for growth mutual funds are summarized in Table 3.8.
Percentage Return

Number of Funds(fj)

Midpoint(mj)

mj fj

(mj X )

(mj X )2

(mj X )2fj

10 but less than 20

20 but less than 30
30 but less than 40
40 but less than 50
50 but less than 60
60 but less than 70
Total

2
9
13
15
5
5
49

15
25
35
45
55
65

30
225
455
675
275
325
1,985

25.51
15.51
5.51
4.49
14.49
24.49

650.7601
240.5601
30.3601
20.1601
209.9601
599.7601

1,301.5202
2,165.0409
394.6813
302.4015
1,049.8005
2,998.8005
8,212.2449

TABLE 3.8 Computations Needed to Calculate the Approximations of the Mean

and Standard Deviation of the 2003 Return for Growth Mutual Funds
Using Equations (3.16) and (3.17) on page 99,
c

X =
X =

mj f j
j =1

n
1,985.0
= 40.51
49
c

and

S =
S =

( m j X )2 f j
j =1

n 1
8,212.2449
49 1

= 171.08843
= 13.08

PROBLEMS FOR SECTION 3.3

Learning the Basics
3.28 Given the following frequency distribution for n = 100:

3.29 Given the following frequency distribution for n = 100:

Class Intervals

Frequency

Class Intervals

Frequency

0Under 10
10Under 20
20Under 30
30Under 40
40Under 50

10
20
40
20
10
100

0Under 10
10Under 20
20Under 30
30Under 40
40Under 50

40
25
15
15
5
100

Approximate
a. the mean.
b. the standard deviation.

Approximate
a. the mean.
b. the standard deviation.

3.4: Exploratory Data Analysis

Applying the Concepts

3.30 A wholesale appliance distributing firm wished to
study its accounts receivable for two successive months.
Two independent samples of 50 accounts were selected for
each of the two months. The results are summarized in the
following table:
Frequency Distributions for Accounts Receivable

Amount

March
Frequency

April
Frequency

6
13
17
10
4
0
50

10
14
13
10
0
3
50

$0 to under $2,000
$2,000 to under $4,000
$4,000 to under $6,000
$6,000 to under $8,000
$8,000 to under $10,000
$10,000 to under $12,000
Total

For each month, approximate the

a. mean.
b. standard deviation.
c. On the basis of (a) and (b), do you think the mean and
the standard deviation of the accounts receivable have
changed substantially from March to April? Explain.
3.31 The following table contains the cumulative frequency distributions and cumulative percentage distributions of braking distance (in feet) at 80 miles per hour for a
sample of 25 U.S.-manufactured automobile models and
for a sample of 72 foreign-made automobile models in a
recent year:
U.S.-Made
Automobile Models
Less Than
Braking
Indicated Values
Distance
(in Ft)
Number Percentage
210
220
230
240
(continued)

0
1
2
3

0.0
4.0
8.0
12.0

3.4

Foreign-Made
Automobile Models
Less Than
Indicated Values
Number Percentage
0
1
4
19

0.0
1.4
5.6
26.4

U.S.-Made
Automobile Models
Less Than
Braking
Indicated Values
Distance
(in Ft)
Number Percentage
250
260
270
280
290
300
310
320

4
8
11
17
21
23
25
25

101

Foreign-Made
Automobile Models
Less Than
Indicated Values
Number Percentage

16.0
32.0
44.0
68.0
84.0
92.0
100.0
100.0

32
54
61
68
68
70
71
72

44.4
75.0
84.7
94.4
94.4
97.2
98.6
100.0

For U.S.- and foreign-made automobiles

a. Construct a frequency distribution for each group.
b. On the basis of the results of (a), approximate the mean
of the braking distance.
c. On the basis of the results of (a), approximate the standard deviation of the braking distance.
d. On the basis of the results of (b) and (c), do U.S.- and
foreign-made automobiles seem to differ in their braking distance? Explain.
3.32 The following data represent the distribution of the
ages of employees within two different divisions of a publishing company.

Age of Employees (Years)

20Under 30
30Under 40
40Under 50
50Under 60
60Under 70

A
Frequency

B
Frequency

8
17
11
8
2

15
32
20
4
0

For each of the two divisions (A and B), approximate the

a. mean.
b. standard deviation.
c. On the basis of the results of (a) and (b), do you think
there are differences in the age distribution between the
two divisions? Explain.

EXPLORATORY DATA ANALYSIS

Section 3.1 discussed sample statistics for numerical data that are measures of central tendency,
variation, and shape. Another way of describing numerical data is thrpough exploratory data
analysis that includes the five-number summary and the box-and-whisker plot (references 5
and 6).

102

CHAPTER THREE Numerical Descriptive Measures

The Five-Number Summary

A five-number summary that consists of
Xsmallest

Q1

Median

Q3

Xlargest

provides a way to determine the shape of the distribution. Table 3.9 explains how the relationships among the five numbers allows you to recognize the shape of a data set.

TABLE 3.9
Relationships among
the Five-Number
Summary and the
Type of Distribution

Type of Distribution
Comparison

Left-Skewed

Symmetric

Right-Skewed

Distance from
Xsmallest to the
median versus the
distance from the
median to Xlargest.

The distance from

Xsmallest to the
median is greater
than the distance
from the median to
Xlargest.

Both distances are

the same.

The distance from

Xsmallest to the
median is less than
the distance from
the median to
Xlargest.

Distance from
Xsmallest to Q1 versus
the distance from Q3
to Xlargest.

The distance from

Xsmallest to Q1 is
greater than the
distance from Q3 to
Xlargest.

Both distances are

the same.

The distance from

Xsmallest to Q1 is less
than the distance
from Q3 to Xlargest.

Distance from Q1 to
the median versus
the distance from
the median to Q3.

The distance from

Q1 to the median is
greater than the
distance from the
median to Q3.

Both distances are

the same.

The distance from

Q1 to the median is
less than the
distance from the
median to Q3.

For the sample of 10 get-ready times, the smallest value is 29 minutes and the largest value
is 52 minutes (see pages 75 and 77). Calculations done previously in section 3.1 show that the
median = 39.5, the first quartile = 35, and the third quartile = 44. Therefore, the five-number
summary is
29 35 39.5 44 52
The distance from the median to Xsmallest to the median (39.5 29 = 10.5) is slightly less
than the distance from Xlargest (52 39.5 = 12.5). The distance from Xsmallest to Q1 (35 29 = 6)
is slightly less than the distance from Q3 to Xlargest (52 44 = 8). Therefore, the get-ready times
are slightly right-skewed.

EXAMPLE 3.15

COMPUTING THE FIVE-NUMBER SUMMARY OF THE 2003 PERCENTAGE RETURN

OF SMALL CAP HIGH-RISK MUTUAL FUNDS
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Compute the five-number summary of the 2003 return for the
small cap mutual funds with high risk. MUTUALFUNDS2004

3.4: Exploratory Data Analysis

103

SOLUTION From previous computations for the 2003 return for the small cap mutual funds
with high risk (see pages 76 and 78), the median = 53.8, the first quartile = 41.7, and the third
quartile = 60.85. In addition, the smallest value in the data set is 37.3 and the largest value is
66.5. Therefore, the five-number summary is
37.3 41.7 53.8 60.85 66.5
The distance from the median to Xlargest (66.5 53.8 = 12.7) is less than the distance
from Xsmallest to the median (53.8 37.3 = 16.5). This indicates left skewness. The distance
from Xsmallest to Q1 (41.7 37.3 = 4.4) is slightly less than the distance from Q3 to Xlargest
(66.5 60.85 = 5.65). This indicates slight right-skewness. Therefore, the results are
inconsistent.

The Box-and-Whisker Plot

A box-and-whisker plot provides a graphical representation of the data based on the fivenumber summary. Figure 3.4 illustrates the box-and-whisker plot for the get-ready times.
FIGURE 3.4
Box-and-Whisker Plot of
the Time to Get Ready

Xsmallest

20

25

30

Q1
35

Median
40
Time (minutes)

Xlargest

Q3
45

50

55

The vertical line drawn within the box represents the median. The vertical line at the
left side of the box represents the location of Q1 and the vertical line at the right side of the
box represents the location of Q3. Thus, the box contains the middle 50% of the values in
the distribution. The lower 25% of the data are represented by a line (i.e., a whisker) connecting the left side of the box to the location of the smallest value, Xsmallest. Similarly, the
upper 25% of the data are represented by a whisker connecting the right side of the box to
Xlargest.
The box-and-whisker plot of the get-ready times in Figure 3.4 indicates very slight rightskewness since the distance between the median and the highest value is slightly more than the
distance between the lowest value and the median. The right whisker is slightly longer than the
left whisker.

EXAMPLE 3.16

THE BOX-AND-WHISKER PLOT OF THE 2003 PERCENTAGE RETURN

OF LOW-RISK, AVERAGE RISK, AND HIGH-RISK MUTUAL FUNDS
The 121 mutual funds that are part of the Using Statistics scenario (see page 72) are classified according to the risk level of the mutual funds (low, average, and high) and type (small
cap, mid cap, and large cap). Construct the box-and-whisker plot of the 2003 return for lowrisk, average-risk, and high-risk mutual funds. MUTUALFUNDS2004

104

CHAPTER THREE Numerical Descriptive Measures

2If

there are outliers, the

whiskers in the Minitab boxand-whisker plot extend to
1.5 times the interquartile
range beyond the quartiles
or to the highest value.

SOLUTION Figure 3.5 is the Minitab box-and-whisker plot of the 2003 return for low-risk,
average-risk, and high-risk mutual funds. Minitab displays the box-and-whisker plot vertically
from bottom (low) to top (high). The asterisk (*) for the average-risk fund represents the presence of outlier values.2 The median percentage return and the quartiles are higher for the highrisk funds than for the low-risk and average-risk funds. The average-risk funds are right-skewed
due to the extremely large return of one fund (78). The high-risk funds appear left-skewed
because of the long lower whisker, but the median return is closer to the first quartile than to
the third quartile. The low-risk funds appear to be slightly right-skewed since the upper whisker
is longer than the lower whisker.

FIGURE 3.5
Minitab Box-andWhisker Plot of the
2003 Return for LowRisk, Average-Risk, and
High-Risk Mutual Funds

Figure 3.6 demonstrates the relationship between the box-and-whisker plot and the polygon for four different types of distributions. (Note: The area under each polygon is split into
quartiles corresponding to the five-number summary for the box-and-whisker plot.)

FIGURE 3.6
Box-and-Whisker Plots
and Corresponding
Polygons for Four
Distributions

Panel A
Bell-shaped distribution

Panel B
Left-skewed distribution

Panel C
Right-skewed distribution

Panel D
Rectangular distribution

3.4: Exploratory Data Analysis

105

Panels A and D of Figure 3.6 are symmetric. In these distributions, the mean and median
are equal. In addition, the length of the left whisker is equal to the length of the right whisker,
and the median line divides the box in half.
Panel B of Figure 3.6 is left-skewed. The few small values distort the mean toward the left tail.
For this left-skewed distribution, the skewness indicates that there is a heavy clustering of values at
the high end of the scale (i.e., the right side); 75% of all values are found between the left edge of
the box (Q1) and the end of the right whisker (Xlargest). Therefore, the long left whisker contains the
smallest 25% of the values, demonstrating the distortion from symmetry in this data set.
Panel C of Figure 3.6 is right-skewed. The concentration of values is on the low end of the
scale (i.e., the left side of the box-and-whisker plot). Here, 75% of all data values are found
between the beginning of the left whisker (Xsmallest) and the right edge of the box (Q3), and the
remaining 25% of the values are dispersed along the long right whisker at the upper end of the scale.

PROBLEMS FOR SECTION 3.4

Learning the Basics
PH Grade
ASSIST

3.33 The following is a set of data from a

sample of n = 5:
7 4 9 8 2

a. List the five-number summary.

b. Construct the box-and-whisker plot and describe the shape.
c. Compare your answer in (b) with that from problem
3.1(d) on page 90. Discuss.
PH Grade
ASSIST

3.34 The following is a set of data from a

sample of n = 6:
7 4 9 7 3 12

a. List the five-number summary.

b. Construct the box-and-whisker plot and describe the shape.
c. Compare your answer in (b) with that from problem
3.2(d) on page 90. Discuss.
PH Grade
ASSIST

3.35 The following is a set of data from a

sample of n = 7:
12 7 4 9 0 7 3

a. List the five-number summary.

b. Construct the box-and-whisker plot and describe the shape.
c. Compare your answer in (b) with that from problem
3.3(d) on page 90. Discuss.
3.36 The following is a set of data from a sample of n = 5:
7 5 8 7 9
a. List the five-number summary.
b. Construct the box-and-whisker plot and describe the shape.
c. Compare your answer in (b) with that from problem
3.4(d) on page 90. Discuss.

Applying the Concepts

Problems 3.373.42 can be solved manually or by
using Microsoft Excel, Minitab, or SPSS.

3.37 A manufacturer of flashlight

batteries took a sample of 13 batteries
from a days production and used them
continuously until they were drained. The number of hours
until failure are in the file. BATTERIES
PH Grade
ASSIST

SELF
Test

342 426 317 545 264 451

1,049 631 512 266 492 562 298
a. List the five-number summary.
b. Construct the box-and-whisker plot and describe the shape.
3.38 In the 20022003 academic year, many public universities in the United States raised tuition and fees due to
a decrease in state subsidies (Mary Beth Marklein, Public
Universities Raise Tuition, Feesand Ire, USA Today,
August 8, 2002, 1A2A). The following represents the
change in the cost of tuition, a shared dormitory room, and
the most popular meal plan from the 20012002 academic
year to the 20022003 academic year for a sample of 10
public universities. COLLEGECOST.XLS
University

Change in Cost ($)

University of California, Berkeley

University of Georgia, Athens
University of Illinois, UrbanaChampaign
Kansas State University, Manhattan
University of Maine, Orono
University of Mississippi, Oxford
University of New Hampshire, Durham
Ohio State University, Columbus
University of South Carolina, Columbia
Utah State University, Logan

1,589
593
1,223
869
423
1,720
708
1,425
922
308

a. List the five-number summary.

b. Construct the box-and-whisker plot and describe the shape.

106

CHAPTER THREE Numerical Descriptive Measures

3.39 A software development and consulting firm located in

the Phoenix metropolitan area develops software for supply
chain management systems using systematic software reuse.
Instead of starting from scratch when writing and developing
new custom software systems, the firm uses a database of
reusable components totaling more than 2,000,000 lines of
code collected from 10 years of continuous reuse effort. Eight
analysts at the firm were asked to estimate the reuse rate
when developing a new software system. The following data
are given as a percentage of the total code written for a software system that is part of the reuse database. REUSE
50 62.5 37.5 75.0 45.0 47.5 15.0 25.0
Source: M. A. Rothenberger, and K. J. Dooley, A Performance
Measure for Software Reuse Projects, Decision Sciences, 30(Fall
1999), 11311153.

a. List the five-number summary.

b. Form the box-and-whisker plot and describe the shape
of the data.
3.40 The following data represent the bounced check fee
(in dollars) for a sample of 23 banks for direct-deposit customers who maintain a $100 balance and the monthly service fee (in dollars) for direct-deposit customers if their
accounts fall below the minimum required balance of
$1500 for a sample of 26 banks. BANKCOST1 BANKCOST2
26 28 20 20 21 22 25 25 18 25 15 20 18 20 25 25 22 30 30 30 15 20 29
12 8 5 5 6 6 10 10 9 7 10 7 7 5 0 10 6 9 12 0 5 10 8 5 5 9
Source: Extracted from The New Face of Banking, Copyright
2000 by Consumers Union of U.S., Inc., Yonkers, NY 107031057.
Adapted with permission from Consumer Reports, June 2000.

a. List the five-number summary of the bounced check fee

and of the monthly service fee.
b. Construct the box-and-whisker plot of the bounced
check fee and the monthly service fee.
c. What similarities and differences are there in the distributions for the bounced check fee and the monthly service fee?
3.41 The following data represent the total fat for burgers
and chicken items from a sample of fast-food chains.
FASTFOOD

3.5

Burgers
19 31 34 35 39 39 43
Chicken
7 9 15 16 16 18 22 25 27 33 39
Source: Extracted from Quick Bites, Copyright 2001 by
Consumers Union of U.S., Inc., Yonkers, NY 107031057. Adapted
with permission from Consumer Reports, March 2001, 46.

a. List the five-number summary for the burgers and for

the chicken items.
b. Construct the box-and-whisker plot for the burgers and
the chicken items, and describe the shape of the distribution for the burgers and chicken items.
c. What similarities and differences are there in the distributions for the burgers and the chicken items?
3.42 A bank branch located in a commercial district of a
city has developed an improved process for serving customers during the noon to 1:00 P.M. lunch period. The waiting time in minutes (operationally defined as the time the
customer enters the line to the time he or reaches the teller
window) of all customers during this hour is recorded over
a period of one week. A random sample of 15 customers is
selected, and the results are as follows: BANK1
4.21 5.55 3.02 5.13 4.77 2.34 3.54
3.20 4.50 6.10 0.38 5.12 6.46 6.19 3.79
Another branch, located in a residential area, is also concerned with the noon to 1 P.M. lunch hour. The waiting time
in minutes (defined as the time the customer enters the line
until he or she reaches the teller window) of all customers
during these hours is recorded over a period of one week. A
random sample of 15 customers is selected, and the results
are as follows: BANK2
9.66 5.90 8.02 5.79 8.73 3.82 8.01
8.35 10.49 6.68 5.64 4.08 6.17 9.91 5.47
a. List the five-number summary of the waiting time at the
two bank branches.
b. Construct the box-and-whisker plot and describe the
shape of the distribution of the two bank branches.
c. What similarities and differences are there in the distribution of the waiting time at the two bank branches?
d. Should you compare the two bank branches? Explain.

THE COVARIANCE AND THE COEFFICIENT OF CORRELATION

In section 2.5, you used scatter diagrams to visually examine the relationship between two
numerical variables. In this section, the covariance and the coefficient of correlation that measure the strength of the relationship between two numerical variables are discussed.

The Covariance
The covariance measures the strength of the linear relationship between two numerical variables
(X and Y). Equation (3.18) defines the sample covariance and Example 3.17 illustrates its use.

3.5: The Covariance and the Coefficient of Correlation

107

THE SAMPLE COVARIANCE

n

cov( X , Y ) =

EXAMPLE 3.17

( X i X )(Yi Y )
i =1

(3.18)

n 1

COMPUTING THE SAMPLE COVARIANCE

Consider the expense ratio and the 2003 return for the small cap high-risk funds. Compute the
sample covariance.
SOLUTION Table 3.10 presents the expense ratio and 2003 return for the small cap high-risk
funds and Figure 3.7 contains a Microsoft Excel worksheet that calculates the covariance for these
data. The Calculations area of Figure 3.7 breaks down Equation (3.18) into a set of smaller calculations. From cell C17, or by using Equation (3.18) directly, the covariance is 1.19738.
cov( X , Y ) =

9.579
9 1

= 1.19738
TABLE 3.10
Expense Ratio and 2003
Return for the Small
Cap High-Risk Funds

FIGURE 3.7
Microsoft Excel
Worksheet for the
Covariance between
Expense Ratio and 2003
Return for the Small
Cap High-Risk Funds

Expense Ratio
1.25
0.72
1.57
1.40
1.33
1.61
1.68
1.42
1.20

2003 Return
37.3
39.2
44.2
44.5
53.8
56.6
59.3
62.4
66.5

108

CHAPTER THREE Numerical Descriptive Measures

The covariance has a major flaw as a measure of the linear relationship between two numerical
variables. Since the covariance can have any value, you are unable to determine the relative
strength of the relationship. To better determine the relative strength of the relationship, you
need to compute the coefficient of correlation.

The Coefficient of Correlation

The coefficient of correlation measures the relative strength of a linear relationship between
two numerical variables. The values of the coefficient of correlation range from 1 for a perfect
negative correlation to +1 for a perfect positive correlation. Perfect means that if the points were
plotted in a scatter diagram, all the points could be connected with a straight line. When dealing
with population data for two numerical variables, the Greek letter is used as the symbol for the
coefficient of correlation. Figure 3.8 illustrates three different types of association between two
variables.
FIGURE 3.8
Types of Association
between Variables

Panel A
Perfect negative
correlation (r = 1)

Panel B
No correlation
(r = 0)

Panel C
Perfect positive
correlation (r = +1)

In panel A of Figure 3.8 there is a perfect negative linear relationship between X and Y.
Thus, the coefficient of correlation equals 1, and when X increases, Y decreases in a
perfectly predictable manner. Panel B shows a situation in which there is no relationship
between X and Y. In this case, the coefficient of correlation equals 0, and as X increases,
there is no tendency for Y to increase or decrease. Panel C illustrates a perfect positive relationship where equals +1. In this case, Y increases in a perfectly predictable manner when
X increases.
When you have sample data, the sample coefficient of correlation r is calculated. When
using sample data, you are unlikely to have a sample coefficient of exactly +1, 0, or 1. Figure
3.9 on page 109 presents scatter diagrams along with their respective sample coefficients of
correlation r for six data sets, each of which contains 100 values of X and Y.
In panel A, the coefficient of correlation r is 0.9. You can see that for small values of
X there is a very strong tendency for Y to be large. Likewise, the large values of X tend to
be paired with small values of Y. The data do not all fall on a straight line, so the association between X and Y cannot be described as perfect. The data in panel B have a coefficient
of correlation equal to 0.6, and the small values of X tend to be paired with large values of
Y. The linear relationship between X and Y in panel B is not as strong as in panel A. Thus,
the coefficient of correlation in panel B is not as negative as in panel A. In panel C the linear relationship between X and Y is very weak, r = 0.3, and there is only a slight tendency
for the small values of X to be paired with the larger values of Y. Panels D through F depict
data sets that have positive coefficients of correlation because small values of X tend to be
paired with small values of Y, and the large values of X tend to be associated with large
values of Y.
In the discussion of Figure 3.9, the relationships were deliberately described as tendencies
and not as cause-and-effect. This wording was used on purpose. Correlation alone cannot prove

3.5: The Covariance and the Coefficient of Correlation

Panel A

Panel B

Panel C

Panel D

Panel E

Panel F

109

FIGURE 3.9 Six Scatter Diagrams Created from Minitab and Their Sample
Coefficients of Correlation r
that there is a causation effect, that is, that the change in the value of one variable caused the
change in the other variable. A strong correlation can be produced simply by chance, by the
effect of a third variable not considered in the calculation of the correlation, or by a cause-andeffect relationship. You would need to perform additional analysis to determine which of these
three situations actually produced the correlation. Therefore, you can say that causation implies
correlation, but correlation alone does not imply causation.
Equation (3.19) defines the sample coefficient of correlation r and Example 3.18 illustrates its use.

110

CHAPTER THREE Numerical Descriptive Measures

THE SAMPLE COEFFICIENT OF CORRELATION

r =

cov( X , Y )
S X SY

(3.19)

( X i X )(Yi Y )
where

cov(X, Y) =

i =1

n 1

( X i X )2
SX =

i =1

n 1

(Yi Y )2
SY =

i =1

n 1

Example 3.18 illustrates the computation of the sample coefficient of correlation using
Equation (3.19).

EXAMPLE 3.18

COMPUTING THE SAMPLE COEFFICIENT OF CORRELATION

Consider the expense ratio and the 2003 return for the small cap high-risk funds. From Figure
3.10 and Equation (3.19), compute the sample coefficient of correlation.
SOLUTION

r =
=

cov( X , Y )
S X SY
1.19738
( 0.287663)(10.554383)

= 0.3943786
FIGURE 3.10
Microsoft Excel
Worksheet for the
Sample Coefficient of
Correlation r between
the Expense Ratio and
the 2003 Return for
Small Cap High-Risk
Funds

3.5: The Covariance and the Coefficient of Correlation

111

The expense ratio and the 2003 return for the small cap high-risk funds are positively correlated. Those mutual funds with the lowest expense ratios tend to be associated with the lowest 2003
returns. Those mutual funds with the highest expense ratios tend to be associated with the highest
2003 returns. This relationship is fairly weak, as indicated by a coefficient of correlation, r = 0.394.
You cannot assume that having a low expense ratio caused the low 2003 return. You can
only say that this is what tended to happen in the sample. As with all investments, past performance does not guarantee future performance.

In summary, the coefficient of correlation indicates the linear relationship, or association,

between two numerical variables. When the coefficient of correlation gets closer to +1 or 1,
the linear relationship between the two variables is stronger. When the coefficient of correlation is near 0, little or no linear relationship exists. The sign of the coefficient of correlation
indicates whether the data are positively correlated (i.e., the larger values of X are typically
paired with the larger values of Y) or negatively correlated (i.e., the larger values of X are typically paired with the smaller values of Y). The existence of a strong correlation does not imply
a causation effect. It only indicates the tendencies present in the data.

PROBLEMS FOR SECTION 3.5

Learning the Basics
3.43 The following is a set of data from a sample of n = 11
items:
X

21

15

24

10

12

15

18

9 18

30

36

12

27

45

54

a. Compute the covariance.

b. Compute the coefficient of correlation.
c. How strong is the relationship between X and Y? Explain.

Applying the Concepts

Problems 3.443.49 can be solved manually or by
using Microsoft Excel, Minitab, or SPSS.

3.44 A recent article (J. Clements, Why Investors Should

Put up to 30% of Their Stock Portfolio in Foreign Funds,
The Wall Street Journal, November 26, 2003, D1) that discussed investment in foreign stocks stated that the coefficient of correlation between the return on investment of
U.S. stocks and International Large Cap stocks was 0.80,
U.S. stocks and International Small Cap stocks was 0.53,
U.S. stocks and International Bonds was 0.03, U.S. stocks
and Emerging market stocks was 0.71, U.S. stocks and
Emerging market debt was 0.58.
a. What conclusions about the strength of the relationship
between the return on investment of U. S. stocks and
these five other types of investments can you make?
b. Compare the results of (a) to those of problem 3.45 (a).
3.45 A recent article (J. Clements, Why Investors
Should Put up to 30% of Their Stock Portfolio in Foreign
Funds, The Wall Street Journal, November 26, 2003, D1)
that discussed investment in foreign bonds stated that the

coefficient of correlation between the return on investment

of U.S. bonds and International Large Cap stocks was
0.13, U.S. bonds and International Small Cap stocks
was 0.18, U.S. bonds and International Bonds was 0.48,
U.S. bonds and Emerging market stocks was 0.20, U.S.
bonds and Emerging market debt was 0.10.
a. What conclusions about the strength of the relationship
between the return on investment of U.S. bonds and
these five other types of investments can you make?
b. Compare the results of (a) to those of problem 3.44 (a).
3.46 The following data COFFEEDRINK represent the calories and fat (in grams) of 16-ounce iced coffee drinks at
Dunkin Donuts and Starbucks:
Product

Calories Fat

Dunkin Donuts Iced Mocha Swirl latte

(whole milk)
Starbucks Coffee Frappuccino blended coffee
Dunkin Donuts Coffee Coolatta (cream)
Starbucks Iced Coffee Mocha Expresso
(whole milk and whipped cream)
Starbucks Mocha Frappuccino blended
coffee (whipped cream)
Starbucks Chocolate Brownie
Frappuccino blended coffee
(whipped cream)
Starbucks Chocolate Frappuccino
Blended Crme (whipped cream)

240
260
350

8.0
3.5
22.0

350

20.0

420

16.0

510

22.0

530

19.0

Source: Extracted from Coffee as Candy at DunkinDonuts and

Starbucks, Copyright 2004 by Consumers Union of U.S., Inc.,
Yonkers, NY 107031057. Adapted with permission from Consumer
Reports, June 2004, 9.

112

CHAPTER THREE Numerical Descriptive Measures

a. Compute the sample covariance.

b. Compute the coefficient of correlation.
c. Which do you think is more valuable in expressing the
relationship between calories and fatthe covariance or
the coefficient of correlation? Explain.
d. What conclusions can you reach about the relationship
between calories and fat?
3.47 The following data represent the value of exports
and imports in 2001 for various countries: EXPIMP
Country

Exports

Imports

874.1
730.8
403.5
266.2
259.9
191.1
158.5
150.4
122.5
121.8

912.8
1180.2
349.1
243.6
227.2
202.0
176.2
141.1
107.3
116.0

European Union
United States
Japan
China
Canada
Hong Kong
Mexico
South Korea
Taiwan
Singapore

3.48 The following data SECURITY represent the

turnover rate of pre-boarding screeners at airports in 19981999 and the security violations
detected per million passengers.
SELF
Test

3.6

Violations

110
100
90
88
79
70
64
53
47
37

20.7
9.9
14.8
25.1
13.5
10.3
13.1
30.1
31.8
14.9

a. Compute the covariance.

b. Compute the coefficient of correlation.
c. What conclusions can you reach about the relationship
between the turnover rate of pre-boarding screeners and
the security violations detected?

a. Compute the covariance.

b. Compute the coefficient of correlation.
c. Which do you think is more valuable in expressing the
relationship between exports and importsthe covariance or the coefficient of correlation? Explain.
d. What conclusions can you reach about the relationship
between exports and imports.

St. Louis
Atlanta
Houston
Boston
Chicago
Denver
Dallas
Baltimore
Seattle/Tacoma

San Francisco
Orlando
WashingtonDulles
Los Angeles
Detroit
San Juan
Miami
New YorkJFK
WashingtonReagan
Honolulu

Turnover

Source: Extracted from Alan B. Krueger, A Small Dose of Common

Sense Would Help Congress Break the Gridlock over Airport
Security, The New York Times, November 15, 2001, C2.

Source: Extracted from N. King and S. Miller, Post-Iraq Influence

of U.S. Faces Test at New Trade Talks, The Wall Street Journal,
September 9, 2003, A1.

City

City

Turnover

Violations

416
375
237
207
200
193
156
155
140

11.9
7.3
10.6
22.9
6.5
15.2
18.2
21.7
31.5

3.49 The following data CELLPHONE represent the digitalmode talk time in hours and the battery capacity in milliampere-hours of cellphones.
Talk
Time

Battery
Capacity

Talk
Time

Battery
Capacity

4.50
4.00
3.00
2.00
2.75
1.75
1.75
2.25
1.75

800
1500
1300
1550
900
875
750
1100
850

1.50
2.25
2.25
3.25
2.25
2.25
2.50
2.25
2.00

450
900
900
900
700
800
800
900
900

Source: Extracted from Service Shortcomings, Copyright 2002 by

Consumers Union of U.S., Inc., Yonkers, NY 107031057. Adapted
with permission from Consumer Reports, February 2002, 25.

a. Compute the covariance.

b. Compute the coefficient of correlation.
c. What conclusions can you reach about the relationship
between the battery capacity and the digital-mode talk
time?
d. You would expect cellphones with higher battery capacity to have a higher talk time. Is this borne out by the
data?

PITFALLS IN NUMERICAL DESCRIPTIVE MEASURES

AND ETHICAL ISSUES
In this chapter you studied how a set of numerical data can be characterized by various statistics that measure the properties of central tendency, variation, and shape. Your next step is
analysis and interpretation of the calculated statistics. Your analysis is objective; your interpre-

Summary

113

tation is subjective. You must avoid errors that may arise either in the objectivity of your analysis or in the subjectivity of your interpretation.
The analysis of the mutual funds based on risk level is objective and reveals several impartial findings. Objectivity in data analysis means reporting the most appropriate descriptive
summary measures for a given data set. Now that you have read the chapter and have become
familiar with various descriptive summary measures and their strengths and weaknesses, how
should you proceed with the objective analysis? Because the data distribute in a slightly asymmetrical manner, shouldnt you report the median in addition to the mean? Doesnt the standard
deviation provide more information about the property of variation than the range? Should you
describe the data set as right-skewed?
On the other hand, data interpretation is subjective. Different people form different conclusions
when interpreting the analytical findings. Everyone sees the world from different perspectives.
Thus, because data interpretation is subjective, you must do it in a fair, neutral, and clear manner.

Ethical Issues
Ethical issues are vitally important to all data analysis. As a daily consumer of information, you
need to question what you read in newspapers and magazines, what you hear on the radio or television, and what you see on the World Wide Web. Over time, much skepticism has been expressed
about the purpose, the focus, and the objectivity of published studies. Perhaps no comment on
this topic is more telling than a quip often attributed to the famous nineteenth-century British
statesman Benjamin Disraeli: There are three kinds of lies: lies, damned lies, and statistics.
Ethical considerations arise when you are deciding what results to include in a report. You
should document both good and bad results. In addition, when making oral presentations and
presenting written reports, you need to give results in a fair, objective, and neutral manner.
Unethical behavior occurs when you willfully choose an inappropriate summary measure (e.g.,
the mean for a very skewed set of data) to distort the facts in order to support a particular position. In addition, unethical behavior occurs when you selectively fail to report pertinent findings because it would be detrimental to the support of a particular position.

SUMMARY
This chapter was about numerical descriptive measures. In
this and the previous chapter, you studied descriptive statisticshow data are presented in tables and charts, and
then summarized, described, analyzed, and interpreted.
When dealing with the mutual fund data, you were able to
present useful information through the use of pie charts,
histograms, and other graphical methods. You explored
characteristics of past performance such as central ten-

TABLE 3.11
Summary of Numerical
Descriptive Measures

dency, variability, and shape using numerical descriptive

measures such as the mean, median, quartiles, range, standard deviation, and coefficient of correlation. Table 3.11
provides a list of the numerical descriptive measures covered in this chapter.
In the next chapter, the basic principles of probability are
presented in order to bridge the gap between the subject of
descriptive statistics and the subject of inferential statistics.

Type of Analysis

Numerical Data

Describing a central tendency, variation,

and shape of a numerical variable

Mean, median, mode, quartiles, geometric

mean, range, interquartile range, standard
deviation, variance, coefficient of variation,
Z scores, box-and-whisker plot (sections
3.13.4)

Describing the relationship

between two numerical variables

Covariance, coefficient of correlation

(section 3.5)

114

CHAPTER THREE Numerical Descriptive Measures

KEY

FORMULAS

Sample Mean

Z Scores

X =

Xi
i =1

(3.1)

(3.12)

Population Mean
N

Median

Median =

n +1
rank value
2

(3.2)

First Quartile Q1

n +1
ranked value
Q1 =
4

Xi
i =1

(3.13)

Population Variance
N

( X i )2

(3.3)

i =1

2 =

Third Quartile Q3

Q3 =

X X
S

Z =

3( n + 1)
ranked value
4

(3.4)

Population Standard Deviation

N

( X i )2

Geometric Mean
1/ n

X G = ( X1 X 2 L X n )

(3.5)

i =1

RG = [(1 + R1 ) (1 + R2 ) L (1 + Rn )]1/ n 1

(3.6)

Approximating the Mean from a Frequency Distribution

c

Range

Range = Xlargest Xsmallest (3.7)

Interquartile Range
Interquartile range = Q3 Q1 (3.8)
Sample Variance
n

X =

mj f j
j =1

Approximating the Standard Deviation from a Frequency

Distribution
c

(3.9)

n 1

( m j X )2 f j
j =1

S =

Sample Covariance

S2 =

(Xi X )
i =1

n 1

(3.10)

Coefficient of Variation

S
CV = 100%
X

KEY

(3.17)

n 1

Sample Standard Deviation

S =

(3.16)

( X i X )2
i =1

(3.15)

Geometric Mean Rate of Return

S2 =

(3.14)

(3.11)

cov( X , Y ) =

( X i X )(Yi Y )
i =1

n 1

(3.18)

Sample Coefficient of Correlation

r =

cov( X , Y )
S X SY

(3.19)

TERMS

arithmetic mean 73
box-and-whisker plot 103
central tendency 72
Chebyshev rule 97
coefficient of correlation 108
coefficient of variation 85
covariance 106
dispersion 72

empirical rule 96
extreme value 86
five-number summary 102
geometric mean 79
interquartile range 81
left-skewed 88
mean 73
median 75

midspread 81
mode 76
outlier 86
population mean 94
population standard deviation
population variance 97
Q1: first quartile 77
Q2: second quartile 77

95

Chapter Review Problems

Q3: third quartile 77

quartiles 77
range 80
resistant measures 81
right-skewed 88
sample coefficient
of correlation 109

CHAPTER

sample covariance 106

sample mean 73
sample standard deviation
sample variance 82
shape 72
skewed 88
spread 72

REVIEW

Checking Your Understanding

3.50 What are the properties of a set of numerical data?
3.51 What is meant by the property of central tendency?
3.52 What are the differences among the mean, median,
and mode, and what are the advantages and disadvantages
of each?
3.53 How do you interpret the first quartile, median, and
third quartile?
3.54 What is meant by the property of variation?
3.55 What does the Z score measure?
3.56 What are the differences among the various measures of variation such as the range, interquartile range,
variance, standard deviation, and coefficient of variation,
and what are the advantages and disadvantages of each?
3.57 How does the empirical rule help explain the ways in
which the values in a set of numerical data cluster and distribute?
3.58 How do the empirical rule and the Chebychev rule
differ?
3.59 What is meant by the property of shape?
3.60 How do the covariance and the coefficient of correlation differ?

Applying the Concepts

You can solve problems 3.613.67 manually or by
using Microsoft Excel, Minitab, or SPSS. We recommend that you solve problems 3.683.86 using
Microsoft Excel, Minitab, or SPSS.

3.61 A quality characteristic of interest for a tea-bag-filling process is the weight of the tea in the individual bags. If
the bags are underfilled, two problems arise. First, customers may not be able to brew the tea to be as strong as
they wish. Second, the company may be in violation of the
truth-in-labeling laws. For this product, the label weight on
the package indicates that, on average, there are 5.5 grams
of tea in a bag. If the average amount of tea in a bag
exceeds the label weight, the company is giving away product. Getting an exact amount of tea in a bag is problematic

115

standard deviation 82
sum of squares 82
symmetrical 88
variance 82
variation 76
Z scores 86

82

PROBLEMS
because of variation in the temperature and humidity inside
the factory, differences in the density of the tea, and the
extremely fast filling operation of the machine (approximately 170 bags a minute). The following table provides
the weight in grams of a sample of 50 tea bags produced in
one hour by a single machine. TEABAGS
5.65
5.57
5.47
5.77
5.61

5.44
5.40
5.40
5.57
5.45

5.42
5.53
5.47
5.42
5.44

5.40
5.54
5.61
5.58
5.25

5.53
5.55
5.53
5.58
5.56

5.34
5.62
5.32
5.50
5.63

5.54
5.56
5.67
5.32
5.50

5.45
5.46
5.29
5.50
5.57

5.52
5.44
5.49
5.53
5.67

5.41
5.51
5.55
5.58
5.36

a. Compute the mean, median, first quartile, and third

quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Interpret the measures of central tendency and variation
within the context of this problem. Why should the company producing the tea bags be concerned about the central tendency and variation?
d. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
e. Is the company meeting the requirement set forth on the
label that, on average, there are 5.5 grams of tea in a bag?
If you were in charge of this process, what changes, if
any, would you try to make concerning the distribution of
weights in the individual bags?
3.62 In New York State, savings banks are permitted to sell
a form of life insurance called Savings Bank Life Insurance
(SBLI). The approval process consists of underwriting, which
includes a review of the application, a medical information
bureau check, possible requests for additional medical information and medical exams, and a policy compilation stage
during which the policy pages are generated and sent to the
bank for delivery. The ability to deliver approved policies to
customers in a timely manner is critical to the profitability of
this service to the bank. During a period of one month, a random sample of 27 approved policies was selected and the following total processing time in days was recorded: INSURANCE
73 19 16 64 28 28 31 90 60 56 31 56 22 18
45 48 17 17 17 91 92 63 50 51 69 16 17

116

CHAPTER THREE Numerical Descriptive Measures

a. Compute the mean, median, first quartile, and third

quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What would you tell a customer who enters the bank to
purchase this type of insurance policy and asks how
long the approval process takes?
3.63 One of the major measures of the quality of service
provided by any organization is the speed with which it
responds to customer complaints. A large family-held
department store selling furniture and flooring, including
carpet, had undergone a major expansion in the past several
years. In particular, the flooring department had expanded
from 2 installation crews to an installation supervisor, a
measurer, and 15 installation crews. A sample of 50 complaints concerning carpet installation was selected during a
recent year. The following data represent the number of
days between the receipt of the complaint and the resolution of the complaint. FURNITURE
54 5 35 137 31 27 152 2 123 81 74 27
11 19 126 110 110 29 61 35 94 31 26 5
12 4 165 32 29 28 29 26 25 1 14 13
13 10
5 27
4 52 30 22 36 26 20 23
33 68
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. On the basis of the results of (a) through (c), if you had
to tell the president of the company how long a customer
should expect to wait to have a complaint resolved, what
would you say? Explain.
3.64 A manufacturing company produces steel housings
for electrical equipment. The main component part of the
housing is a steel trough that is made out of a 14-gauge
steel coil. It is produced using a 250-ton progressive punch
press with a wipe-down operation putting two 90-degree
forms in the flat steel to make the trough. The distance
from one side of the form to the other is critical because of
weatherproofing in outdoor applications. The company
requires that the width of the trough be between 8.31
inches and 8.61 inches. The following are the widths of the
troughs in inches for a sample of n = 49. TROUGH
8.312 8.343 8.317 8.383 8.348 8.410 8.351 8.373 8.481 8.422
8.476 8.382 8.484 8.403 8.414 8.419 8.385 8.465 8.498 8.447
8.436 8.413 8.489 8.414 8.481 8.415 8.479 8.429 8.458 8.462
8.460 8.444 8.429 8.460 8.412 8.420 8.410 8.405 8.323 8.420
8.396 8.447 8.405 8.439 8.411 8.427 8.420 8.498 8.409

a. Calculate the mean, median, range, and standard deviation for the width. Interpret these measures of central
tendency and variability.
b. List the five-number summary.
c. Construct a box-and-whisker plot and describe the shape.
d. What can you conclude about the number of troughs
that will meet the companys requirements of troughs
being between 8.31 and 8.61 inches wide?
3.65 The manufacturing company in problem 3.64 also
produces electric insulators. If the insulators break when in
use, a short-circuit is likely to occur. To test the strength of
the insulators, destructive testing is carried out to determine
how much force is required to break the insulators. Force is
measured by observing how many pounds must be applied
to the insulator before it breaks. The data from 30 insulators
from this experiment are as follows: FORCE
1,870 1,728 1,656 1,610 1,634 1,784 1,522 1,696 1,592 1,662
1,866 1,764 1,734 1,662 1,734 1,774 1,550 1,756 1,762 1,866
1,820 1,744 1,788 1,688 1,810 1,752 1,680 1,810 1,652 1,736

a. Calculate the mean, median, range, and standard deviation for the force variable.
b. Interpret the measures of central tendency and variability in (a).
c. Construct a box-and-whisker plot and describe the shape.
d. What can you conclude about the strength of the insulators if the company requires a force measurement of at
least 1,500 pounds?
3.66 Problems with a telephone line that prevent a customer from receiving or making calls are disconcerting to
both the customer and the telephone company. The following
data represent samples of 20 problems reported to two different offices of a telephone company and the time to clear
these problems (in minutes) from the customers lines:
PHONE

Central Office I Time to Clear Problems (minutes)

1.48 1.75 0.78 2.85 0.52 1.60 4.15 3.97 1.48 3.10
1.02 0.53 0.93 1.60 0.80 1.05 6.32 3.93 5.45 0.97
Central Office II Time to Clear Problems (minutes)
7.55 3.75 0.10 1.10 0.60 0.52 3.30 2.10 0.58 4.02
3.75 0.65 1.92 0.60 1.53 4.23 0.08 1.48 1.65 0.72
For each of the two central office locations:
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a side-by-side box-and-whisker plot. Are the
data skewed? If so, how?
d. On the basis of the results of (a) through (c), are there any
differences between the two central offices? Explain.

Chapter Review Problems

3.67 In many manufacturing processes the term work-inprocess (often abbreviated WIP) is used. In a book manufacturing plant the WIP represents the time it takes for
sheets from a press to be folded, gathered, sewn, tipped on
end sheets, and bound. The following data represent samples of 20 books at each of two production plants and the
processing time (operationally defined as the time in days
from when the books came off the press to when they were
packed in cartons) for these jobs. WIP
Plant A
5.62 5.29 16.25 10.92 11.46 21.62 8.45 8.58 5.41 11.42
11.62 7.29 7.50 7.96 4.42 10.50 7.58 9.29 7.54 8.92

Plant B
9.54 11.46 16.62 12.62 25.75 15.41 14.29 13.13 13.71 10.04
5.75 12.46 9.17 13.21 6.00 2.33 14.25 5.37 6.25 9.71

For each of the two plants:

a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a side-by-side box-and-whisker plot. Are the
data skewed? If so, how?
d. On the basis of the results of (a) through (c), are there
any differences between the two plants? Explain.
3.68 The data contained in the file CEREALS consists of
the cost in dollars per ounce, calories, fiber in grams, and
sugar in grams for 33 breakfast cereals.
Source: Extracted from Copyright 1999 by Consumers Union of
U.S., Inc., Yonkers, NY 107031057. Adapted with permission from
Consumer Reports, October 1999, 3334.

For each variable:

a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you reach concerning the cost per
ounce in cents, calories, fiber in grams, and the sugar in
grams for the 33 breakfast cereals?
3.69 State budget cuts forced a rise in tuition at public
universities during the 20032004 academic year. The data
in the file TUITION include the difference in tuition between
20022003 and 20032004 for in-state students and outof-state students.
a. Compute the mean, median, first quartile, and third
quartile for the difference in tuition between 20022003
and 20032004 for in-state students and out-of-state
students.
b. Compute the range, interquartile range, variance, standard
deviation, and coefficient of variation for the difference in

117

tuition between 20022003 and 20032004 for in-state

students and out-of-state students.
c. Construct a box-and-whisker plot of the difference in
tuition between 20022003 and 20032004 for in-state
students and out-of-state students. Are the data skewed?
If so, how?
d. What conclusions can you reach concerning the difference in tuition between 20022003 and 20032004 for
in-state students and out-of-state students?
3.70 Do marketing promotions, such as bobble-head giveaways, increase attendance at Major League Baseball
games? An article in Sport Marketing Quarterly reported
on the effectiveness of marketing promotions (T. C. Boyd
and T. C. Krehbiel, Promotion Timing in Major League
Baseball and the Stacking Effects of Factors that Increase
Game Attractiveness, Sport Marketing Quarterly,
12(2003), 173183). The data file ROYALS includes the following variables for the Kansas City Royals during the
2002 baseball season:
GAME = Home games in the order they were played
ATTENDANCE = Paid attendance for the game
PROMOTIONY = a promotion was held; N = no promotion was held
a. Calculate the mean and standard deviation of attendance
for the 43 games where promotions were held and for
the 37 games without promotions.
b. Construct a five-number summary for the 43 games
where promotions were held and for the 37 games without promotions.
c. Construct a graphical display containing two boxand-whisker plots; one for the 43 games where promotions were held and one for the 37 games without
promotions.
d. Discuss the results of (a) through (c) and comment on
the effectiveness of promotions at Royals games during
the 2002 season.
3.71 The data contained in the file PETFOOD2 consist of
the cost per serving, cups per can, protein in grams, and fat
in grams for 97 varieties of dry and canned dog and cat
food.
Source: Extracted from Copyright 1998 by Consumers Union of
U.S., Inc., Yonkers, NY 107031057. Adapted with permission from
Consumer Reports, February 1998, 1819.

For the four types of food (dry dog food, canned dog food,
dry cat food and canned cat food), for the variables of cost
per serving, protein in grams, and fat in grams:
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a side-by-side box-and-whisker plot for the
four types (dry dog food, canned dog food, dry cat food,
and canned cat food). Are the data for any of the types of
food skewed? If so, how?

118

CHAPTER THREE Numerical Descriptive Measures

d. What conclusions can you reach concerning any differences among the four types (dry dog food, canned dog
food, dry cat food, and canned cat food)?
3.72 The manufacturer of Boston and Vermont asphalt
shingles provide their customers with a 20-year warranty
on most of their products. To determine whether a shingle
will last as long as the warranty period, accelerated-life
testing is conducted at the manufacturing plant.
Accelerated-life testing exposes the shingle to the stresses
it would be subject to in a lifetime of normal use in a laboratory setting via an experiment that takes only a few minutes to conduct. In this test, a shingle is repeatedly scraped
with a brush for a short period of time and the amount of
shingle granules that are removed by the brushing is
weighed (in grams). Shingles that experience low amounts
of granule loss are expected to last longer in normal use
than shingles that experience high amounts of granule loss.
In this situation, a shingle should experience no more than
0.8 grams of granule loss if it is expected to last the length
of the warranty period. The data file GRANULE contains a
sample of 170 measurements made on the companys
Boston shingles, and 140 measurements made on Vermont
shingles.
a. List the five-number summary for the Boston shingles
and for the Vermont shingles.
b. Construct side-by-side box-and-whisker plots for the
two brands of shingles and describe the shapes of the
distributions.
c. Comment on the shingles ability to achieve a granule
loss of 0.8 grams or less.
3.73 The data in the file STATES represent the results of the
American Community Survey, a sampling of 700,000
households taken in each state during the 2000 U.S.
Census. For each of the variables of average travel-to-work
time in minutes, percentage of homes with eight or more
rooms, median household income, and percentage of mortgage-paying homeowners whose housing costs exceed 30%
of income:
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you reach concerning the average
travel-to-work time in minutes, percentage of homes
with eight or more rooms, median household income,
and percentage of mortgage-paying homeowners whose
housing costs exceed 30% of income?
3.74 The economics of baseball has caused a great deal of
controversy with owners arguing that they are losing money,
players arguing that owners are making money, and fans
complaining about how expensive it is to attend a game and

watch games on cable television. In addition to data related

to team statistics for the 2001 season, the file BB2001 contains
team-by-team statistics on ticket prices; the fan cost index;
regular season gate receipts; local television, radio, and cable
receipts; all other operating revenue, player compensation
and benefits; national and other local expenses; and income
from baseball operations. For each of these variables,
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. Compute the correlation between the number of wins
and player compensation and benefits. How strong is the
relationship between these two variables?
e. What conclusions can you reach concerning the regular
season gate receipts; local television, radio, and cable
receipts; all other operating revenue; player compensation and benefits; national and other local expenses; and
income from baseball operations?
3.75 The data in the file AIRCLEANERS represent the price,
yearly filter energy cost, and yearly filter cost of room air
cleaners.
a. Compute the coefficient of correlation between price
and energy cost.
b. Compute the coefficient of correlation between price
and filter cost.
c. What conclusions about the relationship of energy
cost and filter cost to the price of the air cleaners can
you make?
Source: Extracted from Portable Room Air Cleaners, Copyright
2002 by Consumers Union of U.S., Inc., Yonkers, NY 107031057.
Adapted with permission from Consumer Reports, February 2002, 47.

3.76 The data in the file PRINTERS represent the price, text
speed, text cost, color photo time, and color photo cost of
computer printers.
a. Compute the coefficient of correlation between price and
each of the following: text speed, text cost, color photo
time, and color photo cost.
b. Based on the results of (a), do you think that any of the
other variables might be useful in predicting printer
price? Explain.
Source: Extracted from Printers, Copyright 2002 by Consumers
Union of U.S., Inc., Yonkers, NY 107031057. Adapted with
permission from Consumer Reports, March 2002, 51.

3.77 You want to study characteristics of the model year

2002 automobiles in terms of the following variables: miles
per gallon, length, width, turning circle requirement,
weight, and luggage capacity. AUTO2002
Source: Extracted from The 2002 Cars, Copyright 2002 by
Consumers Union of U.S., Inc., Yonkers, NY 107031057. Adapted
with permission from Consumer Reports, April 2002.

Chapter Review Problems

For each of these variables:

a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you draw concerning the 2002
automobiles?
3.78 Refer to the data of problem 3.77. You want to compare sports utility vehicles (SUVs) with non-SUV vehicles
in terms of miles per gallon, length, width, turning circle
requirement, weight, and luggage capacity. For SUVs and
non-SUV vehicles, for each of these variables:
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a side-by-side box-and-whisker plot. Are the
data skewed? If so, how?
d. What conclusions can you reach concerning differences
between SUVs and non-SUVs?
3.79 Zagats publishes restaurant ratings for various
locations in the United States. The data file RESTRATE
contains the Zagat rating for food, decor, service, and the
price per person for a sample of 50 restaurants located in
New York City, and 50 restaurants located on Long
Island.
Source: Extracted from Zagat Survey 2002 New York City
Restaurants and Zagat Survey 2002 Long Island Restaurants.

50,000

For New York City and Long Island restaurants, for the
variables of food rating, decor rating, service rating, and
price per person:
a. Compute the mean, median, first quartile, and third
quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a side-by-side box-and-whisker plot for the
New York City and Long Island restaurants. Are the data
for any of the variables skewed? If so, how?
d. What conclusions can you reach concerning differences
between New York City and Long Island restaurants?
3.80 As an illustration of the misuse of statistics, an article by Glenn Kramon (Coaxing the Stanford Elephant to
Dance, The New York Times Sunday Business Section,
November 11, 1990) implied that costs at Stanford
Medical Center had been driven up higher than at competing institutions because the former was more likely
than other organizations to treat indigent, Medicare,
Medicaid, sicker, and more complex patients. The chart
below was provided to compare the average 1989 to 1990
hospital charges for three medical procedures (coronary
bypass, simple birth, and hip replacement) at three competing institutions (El Camino, Sequoia, and Stanford).
Suppose you were working in a medical center. Your CEO
knows you are currently taking a course in statistics and calls
you in to discuss this. She tells you that the article was presented in a discussion group setting as part of a meeting of
regional area medical center CEOs last night and that one of
them mentioned that this chart was totally meaningless and
asked her opinion. She now requests that you prepare her
response. You smile, take a deep breath, and reply . . .

What Health Care Costs

A comparison of average 198990 hospital charges in California
for various operations. Sequoia and El Camino Hospitals are
Stanford Medical Centers main local competition.
El Camino

Dollars

40,000

Sequoia
Stanford

30,000
20,000
10,000
0

N/A
Coronary bypass

119

Simple birth

Hip replacement

El Camino costs are the average of high and low charges for a simple
birth with a two-day stay and a hip replacement with a nine-day stay.
Sequoia costs are averages of the middle 50% of all charges for each
operation.
Stanford data are the average cost of all operations.

Source: Stanford Medical Center, Sequoia Hospital, and El Camino Hospital.

120

CHAPTER THREE Numerical Descriptive Measures

3.81 You are planning to study for your statistics examination with a group of classmates, one of whom you particularly want to impress. This individual has volunteered
to use Microsoft Excel, Minitab, or SPSS to get the needed
summary information, tables, and charts for a data set containing several numerical and categorical variables
assigned by the instructor for study purposes. This person
comes over to you with the printout and exclaims, Ive got
it allthe means, the medians, the standard deviations, the
box-and-whisker plots, the pie chartsfor all our variables. The problem is, some of the output looks weird
like the box-and-whisker plots for gender and for major
and the pie charts for grade point index and for height.
Also, I cant understand why Professor Krehbiel said we
cant get the descriptive stats for some of the variablesI
got it for everything! See, the mean for height is 68.23, the
mean for grade point index is 2.76, the mean for gender is
1.50, the mean for major is 4.33. What is your reply?

Report Writing Exercises

3.82 The data found in the data file BEER represent the
price of a six-pack of 12-ounce bottles, the calories per 12
fluid ounces, the percent of alcohol content per 12 fluid
ounces, the type of beer (craft lagers, craft ales, imported
lagers, regular and ice beers, and light and nonalcoholic
beers), and the country of origin (U.S. versus imported) for
each of the 69 beers that were sampled.
Your task is to write a report based on a complete
descriptive evaluation of each of the numerical variables
price, calories, and alcoholic contentregardless of type
of product or origin. Then perform a similar evaluation
comparing each of these numerical variables based on type
of productcraft lagers, craft ales, imported lagers, regular and ice beers, and light or nonalcoholic beers. In addition, perform a similar evaluation comparing and contrasting each of these numerical variables based on the origins
of the beersthose brewed in the United States versus
those that were imported. Appended to your report should
be all appropriate tables, charts, and numerical descriptive
measures.
Source: Extracted from Beers, Copyright 1996 by Consumers
Union of U.S., Inc., Yonkers, NY 107031057. Adapted with
permission from Consumer Reports, June 1996.

TEAM

PROJECTS

The data file MUTUALFUNDS2004 contains information

regarding 12 variables from a sample of 121 mutual funds.
The variables are:
FundThe name of the mutual fund
CategoryType of stocks comprising the mutual
fundsmall cap, mid cap, large cap
ObjectiveObjective of stocks comprising the mutual
fundgrowth or value

AssetsIn millions of dollars

FeesSales charges (no or yes)
Expense ratioratio of expenses to net assets in
percentage
2003 ReturnTwelve-month return in 2003
Three-year returnAnnualized return 20012003
Five-year returnAnnualized return 19992003
RiskRisk-of-loss factor of the mutual fund classified
as low, average, or high
Best quarterBest quarterly performance 19992003
Worst quarterWorst quarterly performance 19992003
3.83 For expense ratio in percentage, 2003 Return, threeyear return, and five-year return,
a. Compute the mean, median, first quartile, and third quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you reach concerning these variables?
3.84 You wish to compare mutual funds that have fees to
those that do not have fees. For each of these two groups,
for the variables expense ratio in percentage, 2003 Return,
three-year return, and five-year return,
a. Compute the mean, median, first quartile, and third quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you reach about differences
between mutual funds that have fees and those that do
not have fees?
3.85 You wish to compare mutual funds that have a
growth objective to those that have value objective. For
each of these two groups, for the variables expense ratio in
percentage, 2003 Return, three-year return, and five-year
return,
a. Compute the mean, median, first quartile, and third quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you reach about differences
between mutual funds that have a growth objective to
those that have value objective?
3.86 You wish to compare small cap, mid cap, and large
cap mutual funds. For each of these three groups, for the
variables expense ratio in percentage, 2003 Return, threeyear return, and five-year return,
a. Compute the mean, median, first quartile, and third quartile.
b. Compute the range, interquartile range, variance, standard deviation, and coefficient of variation.
c. Construct a box-and-whisker plot. Are the data skewed?
If so, how?
d. What conclusions can you reach about differences
between small cap, mid cap, and large cap mutual funds?

Appendix

121

RUNNING CASE
MANAGING THE SPRINGVILLE HERALD
For what variable in the Chapter 2 Managing the
Springville Herald case (see page 62) are numerical
descriptive measures needed? For the variable you identify:
1. Compute the appropriate numerical descriptive measures, and generate a box-and-whisker plot.

WEB

2. Identify another graphical display that might be useful

and construct it. What conclusions can you form from that
plot that cannot be made from the box-and-whisker plot?
Summarize your findings in a report that can be included
with the task forces study.

CASE

Apply your knowledge about the proper use of numerical

descriptive measures in this continuing Web Case from
Chapter 2.
Visit the StockTout Investing Service Web site
www.prenhall.com/Springville/StockToutHome.htm a
second time and reexamine their supporting data and then
answer the following:
1. Reexamine the data you inspected when working on the
Web Case for Chapter 2. Can descriptive measures be
computed for any variables? How would such summary

statistics support StockTouts claims? How would those

summary statistics affect your perception of StockTouts
record?
2. Evaluate the methods StockTout used to summarize the
results of its customer survey www.prenhall.com/
Springville/ST_Survey.htm. Is there anything you
would do differently to summarize these results?
3. Note that the last question of the survey has fewer
responses. What factors may have limited the number
of responses to that question?

REFERENCES
1. Kendall, M. G., and A. Stuart, The Advanced Theory of
Statistics, vol. 1 (London: Charles W. Griffin, 1958).
2. Microsoft Excel 2003 (Redmond, WA: Microsoft
Corporation, 2002).
3. Minitab Version 14 (State College, PA: Minitab Inc.,
2004).
4. SPSS Base 12.0 Brief Guide (Upper Saddle River, NJ:
Prentice Hall, 2003).

5. Tukey, J., Exploratory Data Analysis (Reading, MA:

Addison-Wesley, 1977).
6. Velleman, P. F., and D. C. Hoaglin, Applications, Basics,
and Computing of Exploratory Data Analysis (Boston,
MA: Duxbury Press, 1981).

Appendix 3

Using Software
for Descriptive Statistics
A3.1 MICROSOFT EXCEL
For Descriptive Statistics
Use the Data Analysis ToolPak. Open to the worksheet containing the data you want to summarize. Select Tools 
Data Analysis. From the list that appears in the Data

Analysis dialog box, select Descriptive Statistics and click

OK. In the Descriptive Statistics dialog box (see Figure
A3.1), enter the cell range of the data in the Input Range
box. Choose the Columns option and Labels in First Row
if you are using data that are arranged like the data in the
Excel files on the CD-ROM packaged with this text. Finish

122

CHAPTER THREE Numerical Descriptive Measures

by selecting New Worksheet Ply, Summary statistics,

Kth Largest, and Kth Smallest, and clicking OK. Results
appear on a separate worksheet.
OR you can use any of these sample statistics worksheet
functions in your own formulas including AVERAGE (for
mean), MEDIAN, MODE, QUARTILE, STDEV, VAR,
MIN, MAX, SUM, COUNT, LARGE, or SMALL.

For Coefficient of Correlation

Open the Correlation.xls Excel file, shown in Figure 3.10
on page 110. Follow the onscreen instructions for modifying the table area if you want to use this worksheet with
other pairs of variables. Note in Figure 3.10 that cell E16
contains a formula that uses the COUNT function. This
allows Excel to automatically update the value of n when
the size of the table area is changed and ensures that the
n 1 term is always correct.
The worksheet uses the CORREL function to calculate
the coefficient of correlation. As shown in Figure 3.10, the
formula =E17/(E18 * E19) could also be used in this particular worksheet to calculate the statistic, since the covariance Sx and SY already appear in the worksheet.

A3.2 MINITAB
Computing Descriptive Statistics
To produce descriptive statistics for the 2003 return for different risk levels shown in Figure 3.3 on page 90, open the
MUTUALFUNDS2004.MTW worksheet. Select Stat 
Basic Statistics  Display Descriptive Statistics.
FIGURE A3.1 Data Analysis Descriptive
Statistics Dialog Box
To enter one of these functions into a worksheet, select
an empty cell and then select Insert  Function. In the
Function dialog box, select Statistical from the drop-down
list and then scroll to and select the function you want to
use. Click OK. In the Function Arguments dialog box,
enter the cell range of the data to be summarized and click
OK. (For LARGE and SMALL, enter 1 as the K value; and
for QUARTILE, enter either 1 or 3 as the Quart value, for
either first or third quartile.) In versions of Microsoft Excel
earlier than Excel 2003, you may encounter errors in
results when using the QUARTILE function.

Step 1: In the Display Descriptive Statistics dialog box

(see Figure A3.2), enter C7 or Return 2003 in
the Variables: edit box. Enter C10 or Risk in the
By variables (optional): edit box.

For Box-and-Whisker Plot

See section G.5 (Box-and-Whisker Plot) if you want
PHStat2 to produce a box-and-whisker plot as a Microsoft
Excel chart. (There are no Microsoft Excel commands that
directly produce box-and-whisker plots.)

For Covariance
Open the Covariance.xls Excel file, shown in Figure 3.7 on
page 107. Follow the onscreen instructions for modifying the
table area if you want to use this worksheet with other pairs of
variables. Note in Figure 3.7 that cell C15 contains a formula
that uses the COUNT function. This allows Excel to automatically update the value of n when the size of the table area is
changed, and ensures that the n 1 term is always correct.

FIGURE A3.2 Minitab Display Descriptive

Statistics Dialog Box
Step 2: Select the Statistics button. In the Display
Descriptive StatisticsStatistics dialog box (see
Figure A3.3), select the Mean, Standard deviation,
Coefficient of variation, First quartile, Median,
Third quartile, Interquartile range, Minimum,
Maximum, Range, and N total (the sample size)
check boxes. Click the OK button to return to the

Appendix

123

Display Descriptive Statistics dialog box. Click the

OK button again to compute the descriptive statistics.

FIGURE A3.3 Minitab Display Descriptive Statistics

Statistics Dialog Box

Using Minitab to Create

a Box-and-Whisker Plot
To create a box-and-whisker plot for the 2003 return for
different risk levels shown in Figure 3.5 on page 104,
open the MUTUALFUNDS2004.MTW worksheet. Select
Graph  Boxplot.
Step 1: In the Boxplots dialog box (see Figure A3.4),
select the One Y With Groups choice. (If you
want to create a box-and-whisker plot for one
group, select the One Y Simple choice.) Click the
OK button.

FIGURE A3.5 Minitab BoxplotsOne Y, With Groups

Dialog Box

The output will be similar to Figure 3.5 on page 104.

Calculating a Coefficient of Correlation

To compute the coefficient of correlation for the expense
ratio and the 2003 return for all the mutual funds, open the
MUTUALFUNDS2004.MTW worksheet. Select Stat 
Basic Statistics  Correlation. In the Correlation dialog
box (see Figure A3.6), enter C6 or Expense ratio and C7
or Return 2003. Click the OK button.

FIGURE A3.4 Minitab Boxplots

Dialog Box
Step 2: In the BoxplotOne Y, With Groups dialog box (see
Figure A3.5), enter C7 or Return 2003 in the
Graph variables: edit box. Enter C10 or Risk in the
Categorical variables edit box. Click the OK button.

FIGURE A3.6 Minitab Correlation Dialog Box