5.

MEASURES OF DISPERSION

5.1 Introduction
The values of a variable are generally not all equal. In some cases, the values are very

close to one another; again, in somedifferent from one another. In

cases, they are markedly
order to get a proper idea about the overall nature of a given set of values, it is necessary to
or
know, besides average, the extent to which the given values differ among themselves
distribution
equivalently, how they are scattered about the average. This feature of a frequency is
the values are,
which represents the variability of the given values or reflects how scattered
scatter, 1S
called its dispersion. A device which is used to measure this characteristic, namely
referred to as a measure of dispersion.

5.2 Different measures of dispersion

(a) Range
of its dispersion and it is defined as the
The range of a variable is the simplest measure

and the least of its given set of values.

between the greatest
diference
data are given in a grouped frequency distribution,
the range
noted that if the
It should be lower
the difference between the largest upper boundary and the smallest
can be considered as

boundary.
Tllustration 5.1
takes the values 3, 3, -1, 8, 4, calculate the range
Suppose a variable
value = 8 and
Here the greatest
-1
the least value
=

variable = 8 -(-1) =9.

Thus, the range of the
effect or change or origin and scale on the range of a variahle
Now we shall discuss the

in the following property.

relation Detween two variables x and y, then
If y = a + bx be the
Range ) = 1b| Range (x).

consider that Xmax

and and respectively denote the mavias
x.min respectively maximum and
Let us
of y
minimum of given
values of x, while the corresponding
values are ymay and y.
max and ymin'
we have
Hence, for b > 0,
ymax a t bmax

10 73
STATISTICAL TOOLS AND TECHNIQUES MEASURES OF DISPERSION 75
74

and
Itmay be mentioned that the mean deviation is generally calculated about the arithmetic
min at min
mean. Again, if x1 2 , are the given values of a variable x and f S2 f, are the
so that, Ymax min Omax min corresponding frequencies, then
212602
i.e. Range () =b Range ().
MD,- (Sc)
Again, forb <0, we get
+ where n =
max a
min
and
min a +bmax
$othat, Ymaxmin bmaxmin lustration 5.3
i.e., Range) =-b Range(). .
(ii) oabitedo Compute the mean deviation of a variable assuming the values 3, 5, 9, 1 and 2 about the
Combining () & (i), we get
median.
Here the given values are 1, 2, 3, 5, 9, when arranged in increasing order of magnitude.
Range () = |b | Range ().
Hence median (me) =3.
lustration 5.2 Then using (5a), we have
If two variables x and y are related as 3y + 4x = 9 and range of x is 3, then compute the
range of y.
MDme(1 -

31+ 12-31+13 31+|5 3 + 19 31) =2.2.

Here we have 3y + 4x = 9 or, y =3
lustration 5.4
Find the mean deviation about the mean of following daily wages (in Rs.) of 10 workers:
Range () = xRange )= x 3 = 4. b bee
73, 76, 75, 78, 76, 70, 75, 75, 78, 74

(b) Mean deviation Let x denote daily wages (in Rs.) of workers. From given data first we find the mean ()

The mean deviation is actually the mean of absolute deviations of the given values of the Here 70+73 +74+75 x3+76x2+78 x2-750 - 75.
10
variable from some average. Suppose X, 2 , are the given values of a
variable x and c 10

is achosen Then first considerthe deviations

average. we x-c, (i =1, 2, n)of the values ...
So, mean deviation of x about T is
from The greater the deviations,
c.
is the
more dispersion. To get a suitable measure, it is
to
necessary combine the deviations. The simple arithmetic mean of the deviations will not MD [I70 751 + 173 751 + 74- 75] + 375 751+ 2176 - 751+ 278 751
serve the purpose, since the sum of the deviations may be small even when individually they
are large in magnitude; this is because, deviations of opposite signs cancel each other. So we
take only the magnitudes of the deviations and subseqently calculate the arithmetic mean of T5+2+ 1 +2+= 10 1.6 Rs.)
these absolute deviations. It is termed as the mean deviation of x about c and is denoted by
MD Thus ilustration 5.5
(a)Prove or disprove:i- 25.11=li-25.21
MD k- 5a) i=l

50
In particular, when c = X, mean deviation about mean is We have, i i -

25.11 (25.1 -

) + G-25.1)
i=26

MD, - .(5b) =25 x 25.1 - 25 x 25.1 =

76 STATISTICAL TOOLS AND TECHNIQUES I0MEASsURES OF DISPERSION 77

Hence the mean deviation of x abot mean is

Again. -25.21 25.2-9+ 6-25.2) MD 3l60 621+ 3164 -

62)=2 kg.
25 x252- 25 x 25.2 An important property of mean deviation is as follows
I ftwo variables x and y are related as y = a + bx . G)
then MDOA) = |b| MD)A

where A(x) and A(y) are corresponding values of x and y satisfying the given relation (i).

[Forexample, when A) =
F, A0) =
5 =a + bF]
From ) and i) it is proved that
p-25.11--25.21 Here y a + br. Suppose y takes the value y; when x =
Xp for i = 1, 2, .

(b) Find the mean deviation about A.M. of m, -

(m -), 1, 0, 1, 2, m.
F a + bx, for each i.
Here total number of values = 2m + 1
Again, A) a + bA(x).=

A()) or, b; A) =
|b| lx; A®|
A.M. of the given values Then y A0) blx
-
-

- = -

1 +0 +1 +2+ +(m 1) +m] Hencel; A)] - =

lbl -A), ie., MD)A) =|b| MD()AG)
2m4-1D -m-(m 1) (m-2) n j=l

Tlustration 5,7
Hence, mean deviation about A.M. (= 0) 15 is the relation between two variables x and y and the mean deviation of x
If 2x + 5y of y about its mean.
about its mean 5 is 4, calculate the mean deviation

(2m+1l m -0] +|- (m

-

1)-0+. + 1-1- 0+ p 0 +|1 -0

Here 2x +5y =15 or, y =
3-x
+2 O+. + n O0 16.

+1} +{1 +2 + . +m)]

lence MD,0) =-MD)=x4=
(2m+1[m + (m -1)
+..

lustration 5.8
distribution of age (in years)
of 80 persons is given below:
me12.mim+)) The frequency
Age (in years) Frequency
m(m+) 30-34 12
2m+ 35-39 15
Tlustration 5.6 40-44 18
deviation
64, 60, 60, 64, 60, 64. Calculate the mean
45-49 16
The weights (in kg) of 6 persons are
10
about mean. 50-54
55-59 9
Let x denotes the weight in kg.
Total 80
00X5+64X362
Here mean (T) = 6
Calculate the mean deviation about median.
78 STATISTICAL TOOLS AND TECHNIQUES
MEASURES OF DISPERSION 79
First we construct the following table. Remarks: If the data are given in a frequency table, then we may calculate the mean
TABLE 5.1 deviation about mean using the following simplified form :
NECESSARY CALCULATION FOR MEAN DEVIATION ABoUT MEDIAN
Age Frequency class- cumulative class-mark me ; MD, -T=T-)
xcr
in years) boundaries frequency where n2
30-34 29.5-34.5 12 32 133.32

35-39 15 34.5-39.5 27 91.65 Illustration 5.9

40-44 18 39.5-44.5 45 42 19.98
The scores of a batsman in ten innings during a certain season are
45-49 16 44.5-49.5 61 47 62.24
52 45, 56, 66, 33, 36, 55, 58, 60, 51, 44.
50-54 10 49.5-54.5 71 88.90
55-59 9 54.5-59.5 80 57 125.01 Calculate the mean deviation about mean.
Total 80 521.10 Let x represent the score of the batsman
45+56+..t10 44= =50.4
Here me = 39.5+.
18 x5 =39.5 + 39.5+3.61 =43.11. 10

From the data we find that 4 values of x, namely 45, 33, 36 and 44 are less than .
Hence, M D = = 6.5 years
80 Hence we use the result

deviation discussed below: MD F-) 10((50.4- 45) + (50.4 33) + (50.4 36) + (50.4 - 44))
Some useful results about the mean are

(A) MD-G--7- 64+174 + 14.4 + 6.4)

We know that 43.6 8.72

i x =

---7- Alustration5.10
Find the mean deviation about mean for the following frequency distribution
Now 2-3) = 0 or,
2_-7)+ - 7 ) =0 Height (in inches) Frequency
60-62
a -- ---) 63-65 18
X>X
66-68 42
Hence - =2G- =27-4). 69-71 27
72-74 8

SoS-- --G- First we construct the following table.

ie, MD, G-)= -)

80 STATISTICAL TOOLS AND TECHNIQUES MBASURES OF DISPERSION 81

TABLE 5.2 that,

Without loss of generality we assume

NECESSARY CALCULATION FOR MBAN

Class boundaries be odd and equal 2m 1.

Frequency C) Class mark (x) 67
Case (): Let n to +

So, we have
59.5-62.5 61 -10
62.5-65.5 18 64 -1 -18 S S S Xm SXm+1 S m+ 2 SS2m 2m+1
65.5-68.5 least when x SA S2m+1
42 67 0 0 Now, Ix -Al+ Ix2m+ 1-Al is
68.5-71.5 27 70 27
2-A +I2- A |is least
when
t SASm
71.5-74.5 8 3 16
***

Total 100 SA S *m+2

when
m-AI+ x+2 - A | is least
least when A m+1 =

Here y 15
0.15 andlastly, Ix,+1 A lis in each of the above sums.
O0 lies between the pair of values
Wesee that when A , I i t =

consequently
We consider the values of y which aregreater than y and form the following table. Hence cach of the above sums and
TABLE 5.3 MDA - A I
NECESSARY CALCULATION FOR MEAN DEVIATION
ABOUT MEAN in this case.
S which is the median
cast when A =
x, 1

JO ) Case (ii) : Let n be even and equal to 2m

0.85 27 22.95 Here we have

1.85
Total
14.80
S Smt+1 S. Sm-1 S 2
S. S
37.7 S A S 2m
least when A
o IX Al+ l - Alis
Hence MD, O)- 10-) 100 37.75- 100 0.755 T-A1+ A | is least
when SA
S 2m-1
Sincex 67+3y least when m SA Sm+1*
A | is
1stly, I a-Al+ 1m
-

MD ) =3 x 0.755 = 2.265 (inches) 1

between the pair of values
+
t lies
t between x
and m + I also
(B) The mean deviation in Seen that when A lies above s u m s and consequently
1s least when
measured about median. each the above sums. So each of the
ot
Suppose we are given n values ,
2 of a variable x. Then that
MDA - A I
we are to
prove lies betweer
ot the given values
MDA -A
median
isleast when between x and + 1 Snce
and vhen Alies median also.
is least when A
=

is least when A is the median of the given values. +1 MD

A be the median.
Thus, in eithet is least if
To prove it, we shall use a
simple result
which states case, MD
variable point on the same that, if P and O be
points
line and R be a two
fixcuCS stiration 5.17
between P straight line, then the
listances

of R from P and 18 Ieast When R lies

and o sum ot
a s3. Compute the mean
5.3. Compute
mean deviation about mean and show
1llustration
thatite the data in median.
is r
about
deviation
"Ore than the
mean
STATISTICAL TOOLS AND TECHNIQUES
82 MEASURES OF DISPERSION 83
Here mean of he variable x (say) is

F-8+5+9+ 1 +2) - 4
-
21, since ) = 0and n i=1

Hence
Hence MD,=-(3-4+15 - 41+ 19 - 41+ |l - 41+ 12-4) = 24

From Illustration 5.3, we find

MDne

Hence MD
2.2 (where me is the median)

>
MDme'
- - (5h)

Similarly, on simplification of (5f), we get

In fact, this illustration supports the above result (B).
(C) Standard Deviation 5i)
At the outsetof this section, let us introduce a measure, called the root-mean-square
deviation. It is defined as the positive square root of the arithmetic mean of the . (5j)
the deviations of the variable values from squares o
a chosen
deviation of n values X i 1()n, of variable x about c is
=
average. Thus, the root-mean-square
Ilustration 5.12
- .(5d) Calculate the standard deviation of a variable x which takes the values 1, 3, 5, 7,25.
Here number of values (n) = 13.
Although the
mean-square deviation indicates the scatter of the
obtained for expressing the measure in the same unit as variable, the square-root We calculate
that of the variable.
If we put c= x in (5d), then the measure
is referred to as the
denoted by s. So the standard deviation may be defined standard deviation and =1+3+ 5+. +25 =1+25) =169.
as the positive
of the squares of the deviations of the variable square-root of the meau
values from their mean. Thus 2 = 12 +32 + 52.. + 252 =
(12 +32 ++
26) (22 + 42 ++26)
.5e)
=
26X 2XS5
6 -22(12+22+133) =6201 -4x 15X14x21
6
If the given data are arranged in a
irequency table, since 12 + 22 +32 + +n= n(n+)2n+1)
the standard deviation is 6
given y = 6201 3276 2925

We have s =
where the symbols have their usual significance
It should be noted that s2, the square of the
standard
the variable. deviation, is called the yariance
or, s 13
25-169 56 7.48 (nearly)
To facilitate computation, ( ) nay De
put in a
simpler form. We have Tlustration 5.13
S-3-- =l -3727-72-1G-F) Calculate the standard deviation of weight on the basis of data in
i= Illustration 5.6.