CHAPTER FOUR

MEASUREMENT OF VARIATION /DISPERSION

Specific objectives
At the end of this topic, the trainee should be able to:-
• State the characteristics of a good measure of dispersion.
• Differentiate between the absolute and relative measures.
• Calculate and interpret the measures of dispersion.

Introduction
Measures of variation help us in studying the important characteristics of a
distribution. The measures of dispersion are very useful in statistical work
because they indicate whether the rest of the data are scattered around
the mean or away from the mean. If the data is approximately dispersed
around the mean then the measure of dispersion obtained will be small
therefore indicating that the mean is a good representative of the sample
data. But on the other hand, if the figures are not closely located to the
mean then the measures of dispersion obtained will be relatively big
indicating that the mean does not represent the data sufficiently.

Significance of measuring variation or dispersion

i) to determine the reliability of an average
ii) to serve as a basis for the control of the variability
iii) to compare two or more series with regard to their variability

62
 iv) to facilitate the use of other statistical measures

Property of a good measure of variation/dispersion.

A good measure of variation should possess as far as possible

i) It should be simple to understand

ii) It should be easy to compute
iii) It should be rigidly d3efined
iv) It should be based on every observation of the distribution
v) It should be amenable to further algebraic treatment
vi) It should have sampling stability
vii) It should not be unduly affected by extreme observation

Methods of studying variation

The following Aretha important methods of studying variation

• The range
• The Mean deviation
• The interquartile range or quartile deviation
• The standard deviation
• The Lorenz curve

Absolute and relative measures of variation

Measures of variation may be either absolute or relative. Absolute measures
of variation are expressed in the same statistical unit in which the original
data are given. Such shillings, kilograms, tones, etc
These values may be used to compare the variation in two or more than two
distributions provided the variables are expressed in the same units and
have almost the same average value.
In case the two sets of data are expressed in different units such as
manager’s salary versus workers salary, the absolute measures of variation
are not comparable. In such cases measures of relative variation should be
used.
A measure of relative variation is the ratio of a measure of absolute
variation to an average. It is sometimes called a coefficient of variation,
because coefficient means a pure number that is independent of the unit of
measurement.

a) The Range
The range is defined as the difference between the highest and the smallest
values in a frequency distribution. This measure is not very efficient
because it utilizes only 2 values in a given frequency distribution. However
the smaller the value of the range, the less dispersed the observations are
from the arithmetic mean and vice versa

63
The range is not commonly used in business management because 2 sets of
data may yield the same range but end up having different interpretations
regarding the degree of dispersion.
Range is the simplest method of studying variation. It is defined as the
difference between the value of the smallest observation and the value of
the largest observation

Range = largest value- smallest value

R= L-S

Coefficient of Range = L-S

L+S
Example
The following are the prices of shares of a company from Monday to
Saturday
days Prices
Monday 200
Tuesday 210
Wednesday 208
Thursday 160
Friday 220
Saturday 250

Calculate the range and coefficient of range.

Solution
Range =L-S
L=250
S= 160
Range =250-160
= 90
Coefficient of range = L-S = 250-160 = 90 = 0.219
L+S 250+ 160 410
Merits of range
• It is the simplest to understand and easiest to compute
• It takes minimum time to calculate the value of range

Limitation
• Range is not based on each and every observation of the distribution
• It is subject to fluctuation of considerable magnitude from sample to
sample
• Range cannot be computed in case of open –end distribution
• Range cannot tell us anything about the character of the distribution
within two extreme observations
Uses of range

64
 1. Quality control; the objective of quality control is to keep a check on
the quality of the products without 100% inspection control charts
are prepared.
2. fluctuation in the shares price; range is useful in studying the
variation in the prices in stocks and shares and other commodities
3. Weather forecast; the meteorological department does make use of
the range in determining the difference between the minimum
temperature and maximum temperature.

(b) Mean Deviation

This is a useful measure of dispersion because it makes use of all the values
given. The average deviation is obtained by calculating the absolute
deviation of each observation from median (or mean) and then averaging
this deviation by taking their arithmetic mean. The formula for average
deviation may be written as;
A.D = ∑ /x – med/
N
Incase deviation are taken from mean the formula shall be written as;
A.D x = ∑ /x - x/
N
Computation of average deviation
The formula for computing average deviation is
A.D (med) = ∑ /x – med/
N
If average deviation is small the distribution is highly compact or uniform
since more than half of the cases are concentrated within a small range
around the mean.
The relative measure corresponding to the average deviation called the
coefficient of average deviation is obtained by dividing average by the
measure of central tendency used, i.e. mean or median.

Example 1
In a given exam the scores for 10 students were as follows
Student Mark (x) x−x
A 60 1.8
B 45 16.8
C 75 13.2
D 70 8.2
65
 E 65 3.2
F 40 21.8
G 69 7.2
H 64 2.2
I 50 11.8
J 80 18.2
Total 618 104.4

Required
Determine the absolute mean deviation
618
Mean, x = = 61.8
10

 X-X 104.4
Therefore AMD = = = 10.44
N 10

Example 2
The following data was obtained from a given financial institution. The data
refers to the loans given out in 1996 to several firms

Firms Amount of fx x−x x− x .f

(f) loan per firm
(x)
3 20000 60000 4157.9 12473.70
4 60000 240000 35842.1 143368.40
1 15000 15000 9157.9 9157.9
5 12000 60000 12157.9 60789.50
6 14000 84000 10157.9 60947.40
Σf = 19 Σfx = 286736.90
459000

Required
Calculate the mean deviation for the amount of items given

66
  fx 459, 000
X = = = 24157.9
f 19

 X -X 286736.90
 AMD = =
f 19

= Shs 15, 091.40

NB if the absolute mean deviation is relatively small it implies that the data
is more compact and therefore the arithmetic mean is a fair sample
representative.

(b) The interquartile range or quartile deviation

This is a measure of dispersion which involves the use of quartile. A quartile
is a mark or a value which lies at the boundary of a division when any given
set of data is divided into four equal divisions
Each of such divisions normally carries 25% of all the observations
The semi interquartile range is a good measure of dispersion because it
shows how the rest of the data are generally spread around the mean
The quartiles normally used are three namely;
i. The lower quartile (first quartile Q1) this usually binds the
lower 25% of the data
ii. The median (second quartile Q2)
iii. The upper quartile (third quartile Q3)

Interquartile range = Q3-Q1

Very often the interquartile range is reduced to the form of the semi-
interquartile range or quartile deviation by dividing it by 2
Q.D = quartile deviation
Q.D = Q3-Q1/2
Quartile deviation gives the average amount by which the quartile differs
from the median. Quartile deviation is an absolute measure of variation.
The relative measure corresponding to the measure called the coefficient
of quartile deviation.
Coefficient of Q.D = Q3-Q1/Q3+Q1
The semi-interquartile range,

Q3 - Q1
SIR =
2

Example 1

67
The weights of 15 parcels recorded at the GPO were as follows:
16.2, 17, 20, 25(Q1) 29, 32.2, 35.8, 36.8(Q2) 40, 41, 42, 44(Q3) 49, 52, 55
(in kgs)
Required
Determine the semi interquartile range for the above data
Q3 − Q1 44 - 25 19
SIR = = = = 8.5
2 2 2

Example 2 (Grouped Data)

The following table shows the levels of retirement benefits given to a group
of workers in a given establishment.

Retirement benefits No of retirees UCB cf

£ ‘000 (f)
20 – 29 50 29.5 50
30 – 39 69 39.5 119
40 – 49 70 49.5 189
50 – 59 90 59.5 279
60 – 69 52 69.5 331
70 – 79 40 79.5 371
80 – 89 11 89.5 382

Required
i. Determine the semi interquartile range for the above data
ii. Determine the minimum value for the top ten per cent.(10%)
iii. Determine the maximum value for the lower 40% of the retirees

Solution
The lower quartile (Q1) lies on position
N +1 382 + 1
=
4 4

= 95.75

(95.75 - 50)
 the value of Q1 = 29.5 + x 10
69

68
 = 29.5 + 6.63
= £36.13

The upper quartile (Q3) lies on position

 N + 1
= 3 
 4 
 382 + 1 
=3  
 4 

= 287.25

∴ the value of Q3 = 59.5 +

( 287.25 - 279 ) × 10
52

= 61.08
Q3 - Q1
The semi interquartile range =
2
61.08 - 36.13
=
2
= 12.475
= £12,475
ii. The top 10% is equivalent to the lower 90% of the retirees
The position corresponding to the lower 90%
90
= (n + 1) = 0.9 (382 + 1)
100

= 0.9 x 383

= 344.7

∴ the benefits (value) corresponding to the minimum value for top

10%
= 69.5 +
( 344.7 - 331) x 10
40
= 72.925
= £ 72925

69
 iii. The lower 40% corresponds to position
40
= (382 + 1)
100

= 153.20

∴ Retirement benefits corresponding to its position

= 39.5 +
(153.2 - 119 ) x 10
70
= 39.5 + 4.88
= 44.38
= £ 44380

The 10th – 90th percentile range

This is a measure of dispersion which uses percentile. A percentile is a
value which separates one division from the other when a given data is
divided into 100 equal divisions.
This measure of dispersion is very important when calculating the co-
efficient of skewness.

Example
Using the above data for retirees calculate the 10th - 90th percentile. The
tenth percentile 10th percentile lies on position
10
(382 + 1) = 0.1 x 383
100

= 38.3
∴ the value corresponding to the tenth percentile

(38.3 x 10)
= 19.5 +
50
= 19.5 + 7.66
= 27.16
The 90th percentile lies on position
90
(382 + 1) = 0.9 x 383
100

70
 = 344.7
∴ the value corresponding to the 90th percentile

= 69.5 +
( 344.7 - 331) x 10
40

= 69.5 + 3.425
= 72.925
∴ the required value of the 10th – 90th percentile = 72.925 – 27.16 =
45.765

Merits of quartile deviation

i) In certain respect it is superior to range as a measure of variation
ii) It has a special utility in measuring variation in cases of open –
ended distribution
iii) It has also useful in erratic or highly skewed distribution where
the other measure of variation would be warped by extreme
values. It is not affected by the presence of extreme values.

Limitation
i) Quartile deviation ignores 50% item as the value of quartile
deviation does not depend upon every observation it cannot be
regarded as a god method of measuring variation.
ii) It is not capable of mathematical manipulation
iii) Its value sis very much affected by sampling fluctuation
iv) It is in fact not a measure of variation as it really does not show
the scatter around an average but rather a distance on scale.

(d) Standard deviation

The standard deviation measures the absolute dispersion (or variability of
distribution, the greater amount of dispersion or variability). The greater
the standard deviation the greater will be the magnitude of the deviations
of the value from their mean.
A small standard deviation means a high degree of uniformity of the
observation as well as homogeneity of a series.
A large standard deviation means just the opposite.
Difference between deviation and standard deviation
Both these measure of dispersion are based on each and very item of the
distribution. But they differ in the following respects.
• Algebraic signs are ignores while calculating mean deviation whereas
in the calculation of standard deviation signs are taken into account.

71
 • Mean deviation can be computed either from median or mean. The
standard deviation on the other hand is always computed from the
arithmetic mean because the sum of the squares of the deviation of
items from arithmetic mean is least.
This is one of the most accurate measures of dispersion. It has the following
advantages;
i. It utilizes all the values given
ii. It makes use of both negative and positive values if they occur
iii. The standard deviation reflects an accurate impression of how
much the sample data varies from the mean. This is because its
suitability can also be tested using other statistical methods

Example
A sample comprises of the following observations; 14, 18, 17, 16, 25, 31
Determine the standard deviation of this sample
Observation.

x ( x − x) ( x − x)
2

14 -6.1 37.21
18 -2.1 4.41
17 -3.1 9.61
16 -4.1 16.81
25 4.9 24.01
31 10.9 118.81
Total 121 210.56

121
X= = 20.1
6

( )
2
 x−x 210.56
 Standard deviation,  = =
n 6

= 5.93

Alternative method
x X2
14 196
18 324
17 289
16 256
72
 25 625
31 961
Total 121 2651

2 2
x x 2651  121 
=
2
−  = − 
n  n  6  6 

= 5.93

Example 2
The following table shows the part-time rate per hour of a given no. of
laborers in the month of June 1997.

Rate per hr (x) No. of laborers fx fx2

Shs (f)
230 7 1610 370300
400 6 2400 960000
350 2 700 245000
450 1 450 202500
200 8 1600 320000
150 11 1650 247500
Total 35 8410 2345300

Calculate the standard deviation from the above table showing how the
hourly payment were varying from the respective mean

 fx   fx 
2
2

∴ Standard deviation,  = -
f   f 
 

2
2345300  8410 
= - 
35  35 

= 67008.6 − 577372

= 9271.4

= 96.29

Example 3 – Grouped data

73
In business statistical work we usually encounter a set of grouped data. In
order to determine the standard deviation from such data, we use any of
the three following methods
i. The long method
ii. The shorter method
iii. The coded method
The above methods are used in the following examples

Example 3.1
The quality controller in a given firm had an accurate record of all the iron
bars produced in May 1997. The following data shows those records

i. Using long method

Bar lengths No. of bars(f) Class mid fx fx2
(cm) point (x)
201 – 250 25 225.5 5637.5 1271256.25
251 – 300 36 275.5 9918 2732409
301 – 350 49 325.5 15949.5 5191562.25
351 – 400 80 375.5 30040 11280020
401 – 450 51 425.5 21700.5 9233562.75
451 – 500 42 475.5 19971 9496210.50
501 - 550 30 525.5 15765 8284507.50
313 118981.50 47489526

Calculate the standard deviation of the lengths of the bars

 fx   fx 
2
2

∴ Standard deviation, σ = -
f   f 
 

2
47489526  118981.50 
= - 
313  313 

= 84.99 cm

ii. Using the shorter method

Bar lengths No. of mid point x-A = d fd Fd2

(cm) bars(f) (x)
201 – 250 25 225.5 -150 -3750 562500
251 – 300 36 275.5 -100 -3600 360000
301 – 350 49 325.5 -50 -2450 122500
351 – 400 80 375.5 (A) 0 0 0

74
 401 – 450 51 425.5 50 2550 127500
451 – 500 42 475.5 100 4200 420000
501 - 550 30 525.5 150 4500 675000
Total 313 1450 2267500

Calculate the standard deviation using the shorter method quagmire

 fd   fd 
2
2

∴ Standard deviation, σ = −
f   f 
 

2
2267500  1450 
= − 
313  313 

= 7244.40 − 21.50

= 7222.90

= 84.99 cm

iii. Using coded method

Bar lengths (f) mid point (x) x-A = d d/c = u fu fu2

(cm)
201 – 250 25 225.5 -150 -3 -75 225
251 – 300 36 275.5 -100 -2 -72 144
301 – 350 49 325.5 -50 -1 -49 49
351 – 400 80 375.5 (A) 0 0 0 0
401 – 450 51 425.5 50 1 51 51
451 – 500 42 475.5 100 2 84 168
501 - 550 30 525.5 150 3 90 270
313 29 907

C = 50 where c is an arbitrary number, try picking a

different figure say 45 the answer should be the same.
Standard deviation using the coded method. This is the most preferable
method among the three methods

 fu   fu 
2
2

 = c -
  f 
f  

75
 2
907  29 
= 50  − 
313  313 

= 50 × 1.6997

= 84.99

Variance
Square of the standard deviation is called variance.

(e) Lorenz curve

The Lorenz curve derived by Max O. Lorenz, famous economic statistician,
is a graphic method of studying dispersion. This curve was used by him for
the first time to measure the distribution off wealth and income. Now the
curve is also used to study the distribution of profits, wages, turnover, etc.

Illustration
In the following table is given the number of companies belonging to two
areas A and B according to the amount of profits earned by them. Draw in
the same diagram their Lorenz curve and interpret them.
Profits earned shs (000s) Number of companies
Area A Area B
6 6 2
25 11 38
60 13 52
84 14 28
105 15 38
150 17 26
170 10 12
400 14 4

Solution

76
Relative measures of dispersion
Def: A relative measure of dispersion is a statistical value which may be
used to compare variations in 2 or more samples.
The measures of dispersion are usually expressed as decimals or
percentages and usually they do not have any other units

Example
The average distance covered by vehicles in a motor rally may be given as
2000 km with a standard deviation of 5 km.
In another competition set of vehicles covered 3000 km with a standard
deviation of 10 kms
NB: The 2 standard deviations given above are referred to as absolute
measures of dispersion. These are actual deviations of the measurements
from their respective mean
However, these are not very useful when comparing dispersions among
samples.
Therefore the following measures of dispersion are usually employed in
order to assess the degree of dispersion.
i. Coefficient of mean deviation
Mean deviation
=
mean
ii. Coefficient of quartile deviation
1 (Q - Q )
= 2
3 1

Q2

Where Q1 = first quartile

Q3 = third quartile

iii. Coefficient of standard deviation

Standard deviation
=
mean

iv. Coefficient of variation

77
 standard deviation
= ×100
mean

Example (see information above)

First group of cars: mean = 2000 kms
Standard deviation = 5 kms

∴ C.O.V = 5 x 100
2000

= 0.25%

Second group of cars: mean = 3000 kms

Standard deviation = 10kms
∴ C.O.V = 10 x 100
3000

= 0.33%

Conclusion
Since the coefficient of variation is greater in the 2 nd group, than in the
first group we may conclude that the distances covered in the 1 st group are
much closer to the mean that in the 2nd group.

Example 2
In a given farm located in the UK the average salary of the employees is £
3500 with a standard deviation of £150
The same firm has a local branch in Kenya in which the average salaries are
Kshs 8500 with a standard deviation of Kshs.800
Determine the coefficient of variation in the 2 firms and briefly comment
on the degree of dispersion of the salaries in the 2 firms.
First firm in the UK
C.O.V = 150 x 100
3500

= 4.29%

Second firm in Kenya

C.O.V = 800 x 100

78
 8500

= 9.4%
Conclusively, since 4.29% < 9.4% then the salaries offered by the firm in UK
are much closer to the mean given them in the case to the local branch in
Kenya

COMBINED MEAN AND STANDARD DEVIATION

Sometimes we may need to combine 2 or more samples say A and B. It is
therefore essential to know the new mean and the new standard deviation
of the combination of the samples.

Combined mean
Let m be the combined mean
Let x1 be the mean of first sample
Let x2 be the mean of the second sample
Let n1 be the size of the 1st sample
Let n2 be the size of the 2nd sample
Let s1 be the standard deviation of the 1st sample
Let s2 be the standard deviation of the 2nd sample
n1 x1 + n2 x2
 combined mean =
n1 + n2

n1s12 + n1 ( m − x1 ) + n2 s22 + n2 ( m − x2 )
2 2

combined standard deviation =

n1 + n2

Example
A sample of 40 electric batteries gives a mean life span of 600 hrs with a
standard deviation of 20 hours.
Another sample of 50 electric batteries gives a mean lifespan of 520 hours
with a standard deviation of 30 hours.
If these two samples were combined and used in a given project
simultaneously, determine the combined new mean for the larger sample
and hence determine the combined or pulled standard deviation.

Size x s
40(n1) 600 hrs(x1) 20hrs (s1)
50 (n1) 520 hrs (x2) 30 hrs (s2)

79
 40 ( 600 ) + 50 ( 520 ) 50, 000
Combined mean = = = 555.56
40 + 50 90

Combined standard deviation

40(202 ) + 40(555.56 - 660) 2 + 50(30) 2 + 50( 555.56 - 520)2

=
40 + 50

1600 + 78996.54 + 45000 + 63225.68

=
90

= 47.52 hrs

SKEWNESS
- This is a concept which is commonly used in statistical decision
making. It refers to the degree in which a given frequency curve is
deviating away from the normal distribution
- There are 2 types of skew ness namely
i. Positive skew ness
ii. Negative skew ness

1. Positive Skewness
- This is the tendency of a given frequency curve leaning towards
the left. In a positively skewed distribution, the long tail
extended to the right.

In this distribution one should note the following

i. The mean is usually bigger than the mode and median
ii. The median always occurs between the mode and mean
iii. There are more observations below the mean than above the
mean
This frequency distribution as represented in the skewed distribution curve
is characteristic of the age distributions in the developing countries

80
Frequenc Positively skewed Frequenc
y frequency curve y Negatively skewed
frequency curve

Normal distribution

Long tail
Median
Mode

Mean

Median

Mode
Mean
2. Negative Skewness
This is an asymmetrical curve in which the long tail extends to the left

NB: This frequency curve for the age distribution is characteristic of the
age distribution in developed countries
- The mode is usually bigger than the mean and median
- The median usually occurs in between the mean and mode
- The no. of observations above the mean are usually more than
those below the mean (see the shaded region)

MEASURES OF SKEWNESS
- These are numerical values which assist in evaluating the degree
of deviation of a frequency distribution from the normal
distribution.

- Following are the commonly used measures of skew ness.

1. Coefficient Skewness
= 3
( mean - median )
Standard deviation

2. Coefficient of skewness
mean - mode
=
Standard deviation
NB: These 2 coefficients above are also known as Pearsonian measures of
skewness.

3. Quartile Coefficient of skewness

Q3+ Q1- 2Q2
=
Q3+ Q1
Where Q1 = 1st quartile
Q2 = 2nd quartile

81
 Q3 = 3rd quartile
NB: The Pearsonian coefficients of skewness usually range between –ve 3
and +ve 3. These are extreme value i.e. +ve 3 and –ve 3 which therefore
indicate that a given frequency is negatively skewed and the amount of
skewness is quite high.
Similarly if the coefficient of skewness is +ve it can be concluded that the
amount of skew ness of deviation from the normal distribution is quite high
and also the degree of frequency distribution is positively skewed.

Example
The following information was obtained from an NGO which was giving
small loans to some small scale business enterprises in 1996. the loans are
in the form of thousands of Kshs.

Loans Units Midpoints(x) x-a=d d/c= fu Fu2 UCB cf

(f) u
46 – 50 32 48 -15 -3 -96 288 50.5 32
51 – 55 62 53 -10 -2 -124 248 55.5 94
56 – 60 97 58 -5 -1 -97 97 60.5 191
61 –65 120 63 (A) 0 0 0 0 0 0
66 –70 92 68 5 +1 92 92 70.5 403
71 –75 83 73 10 +2 166 332 75.5 486
76 – 80 52 78 15 +3 156 468 80.5 538
81 – 85 40 83 20 +4 160 640 85.5 57.8
86 – 90 21 88 25 +5 105 525 90.5 599
91 – 95 11 93 30 +6 66 396 95.5 610
Total 610 428 3086

Required
Using the Pearsonian measure of skew ness, calculate the coefficients of
skew ness and hence comment briefly on the nature of the distribution of
the loans.
c (  fu )
Arithmetic mean = Assumed mean +
f

= 63 +
( 428 × 5 )
610
= 66.51

It is very important to note that the method of obtaining arithmetic mean (or any other statistic) by misusing
assumed mean (A) from X and then dividing by c can be abit confusing, if this is the case then just use the
straight forward method of:

Arithmetic mean =
 f .x where x is the midpoint, the answers are the same.
f
82
  fu   fu 
2
2

The standard deviation =c× -

f   f 
 

2
3086  428 
=5 × - 
610  610 

= 10.68
n +1
The Position of the median lies m =
2
610 +1
= = 305.5
2

= 60.5 +
( 305.5 - 191) ×5
120

= 60.5 +
(114.4 ) ×5
120
Median = 65.27
Therefore the Pearsonian coefficient

= 3
( 66.51- 65.27)
10.68
= 0.348

Comment
The coefficient of skewness obtained suggests that the frequency
distribution of the loans given was positively skewed
This is because the coefficient itself is positive. But the skewness is not
very high implying the degree of deviation of the frequency distribution
from the normal distribution is small

Example 2
Using the above data calculate the quartile coefficient of skewness
Q3+ Q1- 2Q2
Quartile coefficient of skewness =
Q3+ Q1

83
 610 +1
The position of Q1 lies on = = 152.75
4

∴ actual value Q1 =55. 5 +

(152.75 - 94 )  5 = 58.53
97

The position of Q3 lies on = 3

( 610 +1) = 458.25
4

∴ actual value Q3 =70.55 +

( 458.25 - 403)
 5 = 73.83 × 5
83

Q2 position: i.e. 2
( 610 +1) = 305.5
4

Actual Q2 value = 60.5 +

( 305.5 -191)  5 = 65.27
120

The required coefficient of skew ness

73.83 + 58.53 − 2 ( 65.27 )
= = 0.013
73.83 + 58.53

Conclusion
Same as above when the Pearsonian coefficient was used

KURTOSIS
84
 -This is a concept, which refers to the degree of peaked ness of a
given frequency distribution. The degree is normally measured
with reference to normal distribution.
- The concept of kurtosis is very useful in decision making
processes i.e. if is a frequency distribution happens to have either
a higher peak or a lower peak, then it should not be used to make
statistical inferences.
- Generally there are 3 types of kurtosis namely;-
i. Leptokurtic
ii. Mesokurtic
iii. Platykurtic
Leptokurtic
a) A frequency distribution which is lepkurtic has generally a
higher peak than that of the normal distribution. The
coefficient of kurtosis when determined will be found to be
more than 3. thus frequency distributions with a value of more
than 3 are definitely leptokurtic
b) Some frequency distributions when plotted may produce a
curve similar to that of the normal distribution. Such
frequency distributions are referred to as mesokurtic. The
degree of kurtosis is usually equal to 3
c) When the frequency curve contacted produces a peak which is
lower that that of a normal distribution when such a curve is
said to be platykurtic. The coefficient of such is usually less
than 3
- It is necessary to calculate the numerical measure of kurtosis.
The commonly used measure of kurtosis is the percentile
coefficient of kurtosis. This coefficient is normally determined
using the following equation
Percentile measure of kurtosis, K (Kappa) = 1
( Q3 - Q1)
2
P90 - P10
Example
Refer to the table above for loans to small business firms/units
Required
Calculate the percentile coefficient of Kurtosis
90
P90 = ( n +1) = 0.9 (610 +1)
100
= 0.9 (611)
= 549.9
The actual loan for a firm in this position
(549.9) = 80.5 +
( 549.9 - 538 ) x 5 = 81.99
40
10
P10 = (n + 1) = 0.1 (611) = 61.1
100
The actual loan value given to the firm on this position is

85
 50.5 +
( 61.1 − 32 ) x 5 = 52.85
62
= 0.9 (611)
= 549.9
∴ Percentile measure of kurtosis

K (Kappa) = ½
( Q3 - Q1)
P90 - P10

=½
( 73.83 - 58.53)
81.99 - 52.85

= 0.26
Since 0.26 < 3, it can be concluded that the frequency distribution
exhibited by the distribution of loans is platykurtic
Kurtosis is also measured by moment statistics, which utilize the exact
value of each observation.

i. M1 the first moment = M1 =

 X = Mean M1 or M1
n

M2 =
X 2

M3 =
X 3

M4 =
X 4

3. M2 second moment about the mean M2 or f2

M2 = M2 – M12
4. M3 third moment about the mean M3 (a measure of the absolute
skew ness)
M3 = M3 – 3M2M1 + 2M13
5. M4 fourth moment about the mean M4 (a measure of the absolute
Kurtosis)
M4 = M4– 4M3M1 + 6M2M12 + 3M14
An alternative formula
 ( x − m) f
4

M4 = Where m is mean
f

86