Measures Of

Dispersion
MEANING OF DISPERSION
The essential purpose of statistical averages discussed in the preceding
chapter is to summarize a large mass of data. These averages serve to
locate the ‘center’ of a distribution but they do not reveal how the items
or the observations are spread out or scattered around their mean. This
latter characteristic of a distribution is variously known as the dispersion,
‘scatter’ or ‘variation’. If the dispersion is small, it indicates high uniformity
of the observations in the distribution. Absence of dispersion in the data
indicates perfect uniformity. This situation arises when all observations in
the distribution are identical.
MEANING OF DISPERSION
The study of dispersion bears its importance from the fact that various distributions may
have exactly the same averages, but substantial differences in variability. We illustrate this
point by the following example:
Suppose that two students secured the following marks in an examination:

Math Statistics Geography English Average

Student 1 49 50 55 54 52
Student 2 68 10 90 40 52

The two distributions are certainly not identical though their averages are the same. These
differences lie in the dispersion of their marks. The student 1 shows little variation in his
secured marks, while the second student has much greater variation in the marks he
secured.
MEASURES OF DISPERSION
The measures of dispersion can broadly be classified to fall into one of the two
categories: absolute measures and relative measures. The first category of
measures (absolute measures) includes:
I. The range
II. The quartile deviation
III. The mean (or average) deviation
IV. The variance
V. The standard deviation
The measures are absolute in the sense that they are expressed in the same
statistical unit in which the original data are presented, such as dollar, taka, meter,
kilogram, etc.
MEASURES OF DISPERSION
When the two or more data sets are expressed in different units, however, the
absolute measures are not comparable, in which case it is necessary to
consider some other measures that reduce the absolute deviation in some
relative form. These measures are referred to as relative measures. The relative
measures are usually expressed in the form of coefficients and are pure
numbers, independent of the unit of measurements. The measures are
I. Coefficient of range
II. Coefficient of quartile deviation
III. Coefficient of mean deviation
IV. Coefficient of variation
ABSOLUTE MEASURES OF DISPERSION
The Range
The simplest and the crudest measure of dispersion is the range
(R). This is defined as the difference between the smallest (S) and
the largest (L) values in the distribution.
R=L–S
For a set of observations 90, 110, 20, 51, 210 and 190, say, the smallest
value is 20 and the largest value is 210, so that R = 210  20 = 190.
Mean Deviation

For data clustered near the central value, the differences of the
individual observations from their typical value will tend to be small.
Accordingly, to obtain a measure of the total variation in the data, it is
appropriate to find an average of these differences. The resulting
average will be called mean deviation. It is also known as the average
deviation.
Mean Deviation
In practice, the mean deviation is computed as the arithmetic mean of the
absolute values of the deviations from a typical value of a distribution. The
typical value may be the arithmetic mean, median, mode or any other
arbitrary (any value without mean) value. The median is sometimes
preferred as a typical value, because the sum of the absolute values of the
deviations from the median is smaller than any other value. In practice,
however, the arithmetic mean is generally used.
 Mean Deviation
If x1, x2,..,xn form a sample of observations, the formula for computing the
average or mean deviation about any arbitrary values ‘a’ is

M d a  
x i a
… (4.4)
n
where  means that the signs of the deviations whether positive or
negative, are ignored. For a grouped frequency distribution with fi=n, the
mean deviation about the arbitrary value ‘a’ is

Md (a) 
 f i xi  a
… (4.5)
n
If we replace ‘a’ by x , the resulting mean deviation will be called mean
deviation about the mean:

M d x  
x i x
… (4.5a)
n
where n= f i .For a grouped frequency distribution
 n
If we replace ‘a’ by x , the resulting mean deviation will be called mean
Mean Deviation
deviation about the mean:

M d x  
x i x
… (4.5a)
n
For a grouped frequency distribution, where 𝑛 =
 𝑖 𝑓,
where n= f i .For a grouped frequency distribution

M d x  
f i xi  x
… (4.5b)
n
When the deviations are taken from the median we substitute m ~ for a in
(4.5), and the resulting formula for computing mean deviation about the
median is

M d m
~  f i
~
xi  m
… (4.5c)
n
The following examples demonstrate how the mean deviation is computed.
Mean Deviation
Steps to compute mean deviation:
To compute the mean deviation about mean for the given data, the
following steps are involved:
a) Compute the arithmetic mean. This is 127.6 in the present instance.
b) Obtain the absolute deviation of each value in column (2) of Table 4.1
from the computed mean. These deviations are shown in column (3).
c) Obtain the sum of column (3) and divide the resulting sum by the
total number of observations (n=10).
d) The result obtained in (c) above is the mean deviation about the
mean.
Mean Deviation
Example: Ten persons of varying ages were weighed and the following
weights in kg were recorded:

110, 125, 125, 147, 117, 125, 136, 157, 124, 110.

Compute mean deviation about the mean, median and an arbitrary value
120.

Solution: Repeat the procedure outlined above to compute the mean

deviations about the median (which is 125 for the data set) and the
arbitrary value 120, i.e. a=120. The corresponding deviations are shown in last
two columns of Table 4.1.
Mean Deviation
Table 4.1: Computation of mean deviations
Serial no. Weight (xi) xi  x ~
xi  m xi  a
(1) (2) (3) (4) (5)
1 110 17.6 15.0 10.0
2 125 2.6 0 5.0
3 125 2.6 0 5.0
4 147 19.4 22.0 27.0
5 117 10.6 8.0 3.0
6 125 2.6 0 5.0
7 136 8.4 11.0 16.0
8 157 29.4 32.0 37.0
9 124 3.6 1.0 4.0
10 110 17.6 15.0 10.0
Total 1276 114.4 104.0 122.0

The mean deviations about the mean, median and an arbitrary value 120
 7 136 8.4 11.0 16.0
8 157 29.4 32.0 37.0

Mean Deviation
9 124 3.6 1.0 4.0
10 110 17.6 15.0 10.0
Total 1276 114.4 104.0 122.0

The mean deviations about the mean, median and an arbitrary value 120
are respectively

Md (x) 
 x i x

 x i  127.6

114.4
 11.44
n n 10

Md ~
( m) 
 x i
~
m

 x i  125

104.0
 10.40
n n 10
and

Md (a) 
 x i a

 x
122.0
i  120
 12.20 
n n 10
Note that among the three mean deviations, mean deviation about the median
is the smallest.
Variance and Standard Deviation
Instead of ignoring the signs of deviations from the mean as in the
computation of an average deviation, they may each be squared and then
the results are added. The sum of squares can be regarded as a measure of
the total dispersion of the distribution. By dividing the sum by n (the total
number of observations), we obtain the average of the squares of deviations,
a measure, called variance, of the distribution. If the observations are all from
a population, the resulting variance is referred to as the population variance.
As a formula, the variance of population observations 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑁 , commonly
designated 𝜎 2 is
2
(𝑥𝑖 − 𝜇)2
𝜎 =
𝑁
 Variance and Standard Deviation
where 𝜇 is the mean of all the observations in the population and N is the total number
of observations in the population. Because of the operation of squaring, the variance is
expressed in square units (e.g. km2, taka2, etc.), and not (e.g. km, taka, etc.), of the
original unit. It is therefore necessary to extract the positive square root to restore the
original unit. The measure of dispersion thus obtained is called the population standard
deviation and is usually denoted by 𝜎. Thus

(𝑥𝑖 − 𝜇)2
𝜎=
𝑁

Thus, by definition, the standard deviation is the positive square root of the mean-
square deviations of the observations from their arithmetic mean.
Variance and Standard Deviation
Fortunately, it can be shown that if the sum of the squared deviations in the
sample is divided by n–1, and not by n, then the resulting sample variance will
provide an unbiased estimate of the population variance. For this reason, the
sample variance is defined as follows:

2
(𝑥𝑖 − 𝑥)2
𝑠 =
𝑛−1

Such an estimate will show no systematic tendency to be either greater than

or less than the population variance 𝜎 2 . The division by n–1 instead of n
makes the average squared deviation consistent with many similar
measures used in statistical measures.
Computing Variance for Ungrouped Data
The variance and hence the standard deviation are simple to compute
for ungrouped data. Suppose a data set consists of n values 𝑥1 , 𝑥2 , ⋯ ,
𝑥𝑛 . As a first step, compute the arithmetic mean for this data set. Then
subtract this mean from each of the values of x and obtain a set of
deviations 𝑥1 − 𝑥 , (𝑥2 −𝑥), ⋯ , (𝑥𝑛 −𝑥). Then square these deviations, sum
them and divide the resulting sum by n. This gives you the variance of
the given values 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑛 . Let us illustrate the computation of
variance from raw data by an example.
Computing Variance for Ungrouped Data

Example: Compute the variance and standard deviation from the

data on weight of ten children in Example 4.1.

Solution: The data were as follows: 20, 13, 17, 17, 13, 18, 14, 17, 16, and 15.
The mean of this set is 16. Following the steps outlined above, the
accompanying table is constructed to illustrate the computation of
variance.
Computing Variance for Ungrouped Data
Table 4.3: Computation of variance and standard deviation
Child xi xi  x ( xi  x ) 2 x2
1 20 4 16 400
2 13 –3 9 169
3 17 1 1 289
4 17 1 1 289
5 13 –3 9 169
6 18 2 4 324
7 14 –2 4 196
8 17 1 1 289
9 16 0 0 256
10 15 –1 1 225
Total 160 0 46 2606

The variance is thus

s 2

 (x i  x )2

46
 5.11 kg
2
n 1 9
Taking square root of the variance, we obtain the standard deviation:
s  5.11 kg 2  2.26 kg
Computing Variance for Frequency Distribution

The formula for computation of variance and standard deviation of a

frequency distribution should be modified to take into account the values
of x and their corresponding frequencies. Thus, if the variable values
𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘 each occur with frequencies 𝑓1 , 𝑓2 , ⋯ , 𝑓𝑘 respectively, then

2
𝑓𝑖 (𝑥𝑖 − 𝑥)2
𝑠 =
𝑛−1

For grouped data 𝑥𝑖 will be the mid-value of the i-th class.

Computing Variance for Frequency Distribution
The formula for the computation of the variance presented above can be
rewritten in a compact form as follows:

1   f i xi 2 
s 
2
n 1  
 f i xi2 
n


 



n f i xi2   
f i xi 2
… (4.15)
nn  1

In many textbooks, the divisor n is used in place of n1. The discrepancy in

the value of s2 resulting from the use of n instead of n-1 is not however
substantial when n is large.
Computing Variance for Frequency Distribution
Example 4.6: Compute the variance and standard deviation for the
following frequency distribution:

x: 3 5 7 8 9
f: 2 3 2 2 1

Solution: The following table illustrates the computation of variance from

the above distribution.
xi fi f i xi f i xi2
3 2 6 18
5 3 15 75
7 2 14 98
8 2 16 128
9 1 9 81
Total 10 60 400

n  f i x i2   f x  i i
2
10400  602
s 
2
  4.44
nn  1 1010  1
Computing Variance for Frequency Distribution
Example 4.7: The lengths of 32 leaves were measured correct to the
nearest mm. Find the mean, variance and hence the standard deviation of
the lengths.

Length: 20-22 23-25 26-28 29-31 32-34

Frequency: 3 6 12 9 2

Solution: In order to compute the required measures, we construct the

following table:

Length xi xi2 fi f i xi f i xi2

20–22 21 441 3 63 1323
23–25 24 576 6 144 3456
26–28 27 729 12 324 8748
29–31 30 900 9 270 8100
32–34 33 1089 2 66 2178
Total - - 32 867 23805

The mean length is

fx i i 867
 i i i i i i i
20–22 21 441 3 63 1323
23–25 24 576 6 144 3456
Computing Variance for Frequency Distribution
26–28
29–31
27
30
729
900
12
9
324
270
8748
8100
32–34 33 1089 2 66 2178
Total - - 32 867 23805

The mean length is

x
fx i i

867
 27.1
n 32
Using (4.15) the variance is
n f i xi 2   f i xi 2 32(23805)  867 2
s 
2
  10.15
n(n  1) 32  31

so that the standard deviation is s  10 .15  3.19

Hence the mean, variance and standard deviation are 27.1, 10.15 and 3.19
respectively.
Properties of Variance
4.1: The variance is independent of origin but dependent on the scale of
measurement.
4.2: If u=x+y,then su2  s x2  s 2y  2 cov x, y  , where su2 is the variance of u
and Cov(x, y) is the covariance between x and y as defined below:

Covx, y  
 x i  x  yi  y 
n

4.3: The variance is minimized if computed from the arithmetic mean

 Uses of Standard Deviation
A thorough understanding of the use of standard deviation is difficult for us
at this stage, unless we acquire some knowledge on some theoretical
distributions in statistics. Nevertheless, we shall try to introduce the idea of its
use through a few simple illustrative examples. The standard deviation of a
population (𝜎) is a measure of the dispersion in the population, while the
standard deviation of sample observations (s) is a measure of the
dispersion in the distribution constructed from the sample. In both the cases,
the standard deviation (like the mean deviation) represents the average
variability in a distribution. The greater this variability around the mean of a
distribution, the larger the standard deviation. Thus s = 4.5, for example,
indicates greater variability than s = 2.5.
RELATIVE MEASURES OF DISPERSION
Coefficient of Variation
The coefficient of variation (CV) is one of the important measures of
dispersion that attempts to measure the variability in data relative to
the mean. When mean values of two or more data sets vary
considerably, we do not get an accurate picture of the relative
variability in the sets just by comparing the standard deviations.
Coefficient of variation tends to overcome this difficulty. This is a
measure that represents the spread of the distribution relative to the
mean of the same distribution.
RELATIVE MEASURES OF DISPERSION
A coefficient of variation is computed as a ratio of the standard deviation
of the distribution to the mean of the same distribution. Expressing in
percentage form, the symbolic representation of the coefficient is:
𝑠𝑥
𝐶𝑉 = 𝑥
× 100
Clearly, if the mean of a data set is zero, CV cannot be computed. The
measure is a pure number and independent of units.
A value of 33 percent, for example, for CV implies that the standard
deviation of the sample value is 33 percent of the mean of the same
distribution.
RELATIVE MEASURES OF DISPERSION

As an illustration of the use of CV as descriptive statistics, let us

look at the following example:

Example 4.17: The average weekly wage in a factory had increased

from Tk.8000 to Tk.12000 as result of negotiation between the
employees and the employer. Alongside, the standard deviation
had decreased from Tk.150 to Tk.100. Can we conclude that after
negotiation, the wage has become higher and more uniform?
RELATIVE MEASURES OF DISPERSION
Solution: As the standard deviation after the settlement shows a lower
value than before, one might tend to conclude that disparity in wage has
been considerably reduced. But the average wage differs considerably
before and after the settlement. It is therefore not safe to base our
decision only on the basis of standard deviation. Coefficient of variation
seems to be the best tool in this instance. Thus
RELATIVE MEASURES OF DISPERSION
Solution:
Thank You