UNIT 4 MEASURES OF CE-NTRAL

TENDENCY
Structure
Objectives
Introduction
Measures of Central Tendency
4.2.1 Arithmetic Mean
4.2.2 Median
4.2.3 Mode
Other Measures of Central Tendency
4.3.1 Geometric Mean and Harmonic Mean
4.3.2 Weighted Mean
4.3.3 Pooled Mean
4.3.4 Choosing a Measure of Central Tendency
Percentiles
4.4.1 Percentiles: Definition and Computation
4.4.2 Quartiles and Deciles
Let Us sum Up
Key Words
Some Useful Books
Answers or Hints to Check Your Progress Exercises

4.0 OBJECTIVES
After going through this unit, you will be able to:
compute numerical quantities that measure the central tendency of a set of data
such as, mean, median, mode, geometric mean and @manic mean, and
use these measures.

4.1 INTRODUCTION
In the previous Unit we had distxssed about condensation of raw data by grouping
them into a few class intervals and presenting in the form of a table or diagram.
Such tables or diagrams provide a rough idea of the distribution of observations.
OAen we need to compare between distributions. In such situations it is difficult
to compare tables or diagrams simply by looking at them. It is much more
convenient and usefbl for comparison if we could find out a single numerical value
for describing the data.

Measures of Central Tendency (or Location) constitute one of the major statistics
designed for this purpose. There are five main measures of central tendency. These
are Arithmetic Mean, Geometric Mean, Harmonic Mean, Median and Mode. You
will learn about each one of these measures below.
S u n ~ n ~ n r i s a t i o onf
Univariate Data 4.2 MEASURES OF CENTRAL TENDENCY
In frequency distributions of observations discussed in Unit 3 we notice that
the observations tend to cluster around a central value. This phenomenon of
clustering around a central value in a frequency distribution is called 'Central
Tendency'. Thus, it is of interest to locate such a value around which clustering
of observations takes place. There are several measures of central tendency (or
location) of a frequency distribution. These measures produce numbers that
summarise a frequency distribution in terms of one of its properties, namely, central
tendency.
4.2.1 Arithmetic Mean
The average or the arithmetic mean, or simply the mean when there is no
ambiguity, is the most common measure of central tendency. It is defined as the
sum total of all values in the sample divided by the number of observations. It is
denbted by a bar above the symbol of the variable being averaged. Thus stands
for the mean,of X-values in the sample. If in a sample a particular X-value, say
X occurs with frequency4 (i = 1,2, ... n), its contribution to the total of X-values
isJlX,. Thus, one can compute the mean of X-values by
n

When observations are classified into class intervals, as for continuous variables,
individual observations falling into a class interval are not separately identifiable
and the contribution of the individual observation from a class interval to the total
cannot be calculated. To avoid this difficulty, it is assumed that every observation
falling into a class interval has a value equal to the mid-point into which these
observations fall. Such a procedure will not give the exact mean had one
computed it from raw data and may require what is called corrections for
grouping.

Example 4.1: Compute the mean for discrete frequency distribution of Table 4.1.

Table 4.1
Frequency distribution of 100 households by size

Household Size (XJ Frequency (x

1 3
2 16

3 25
4 33
5 12

6 7
7 2
8 2
Total 100
 Let us compute the arithmetic mean of the data given in the above table. Measures o f Central
Tendency

Thus, mean household size based on 100 households is 3.74.

Example 4.2: Compute the-meanfor grouped frequency distribution of Table 4.2.

Table 4.2
Frequency distribution of 100 households by average monthly household
expenditure on food
-
Expenditure class (Rs.) Frequency

262.5 - 286.5 1
286.5 - 310.5 14
310.5 - 334.5 16
334.5- 358.5 28
358.5 - 382.5 26
382.5- 406.5 15

Total 100

For computation of the mean we have to construct table as given below.

I

Class interval (Rs.) Mid-point (3) Frequency Cf;) 44

(0) (1) (2) (3)
262.5 - 286.5 274.5 1 274.5
286.5 - 310.5 298.5 14 4179.0
310.5- 334.5 322.5 16 5160.0
334.5- 358.5 346.5 28 9702.0
358.5 - 382.5 370.5 26 W3.0
382.5- 406.5 394.5 15 5917.5

Total 100 "\:: 34866.0

Thus, mean of monthly average household expenditure on food is
-
. .
,x=--
34866 - Rs. 348.66.
100

One may note from the above example that to find column (3) one needs to multiply
the corresponding values of column (1) and (2), and often hand computations are
long for each multiplication. Thesecomputations can be simplified, particularly when
successive column (1) values are equidistant (but applicable otherwise alsb), by
making the following simple transforrnatio~ .
For i = 1, 2, ... , n

X i- A
u. = -
k i.e., 4.= A + hu, . and so X = A + ~ Z .
Summarisation o f
Univariate Data
Often A is called the 'assumed mean' and hc as its correction to get X . Choice
of A and h are made so that computation of ii becomes simple. Usually A is taken
as that X value for which the frequency is largest. For equidistant successive'
X-values in column (I), h may be taken as the difference between two successive
X-values. For equal length class intervals, the difference between suiccessive
mid-points is the same as the length of each class interval.

We will explain this method by re-computing the mean of the monthly average
household food expenditure data given in Table 4.2. We construct Table 4.3 by
using A and h as explained below.

We define A = Mid-point of the class with largest frequency 346.5 and

h = Common length of each class interval = 24.
-

Table 4;3
Computation of mean of data of Table 4.2

X,- 346.5
Class interval Mid-point u, = Frequency f ; Ut
24
(Rs.1 <q u;)
262.5 - 286.5 274.5 - 3 1 -3
286.5- 310.5 298.5 -2 14 - 28
310.5 - 334.5 , 322.5 -1 16 - 16
334.5- 358.5 346.5 0 28 0
358.5 - 382.5 370.5 1 26 26
382.5- 406.5 394.5 2 15 30

Total 100
9J
We find out that

Properties of Arithmetic Mean

1) The algebraic sum of deviations of a given set of observations is zero

when taken from the arithmetic mean.
Let X I ,X,, ..... Xn be n observations with respective frequencies as f;,

f,, .....f.. Mathematically, this property implies that i-I

f,h (xi - E)= 0,

where xi- H is the deviation of zth observation from mean.

To prove the above property, we write
 Measures o f Central
Tendency

Hence the result.

2) The sum of squares of deviations of a given set of observations is
minimum when taken from the arithmetic mean.
Mathematically, this property implies that for any arbitrarily chosen origin, A,
s= t/;(xi - AY is minimum when A = j7.

To prove this property, we note that the magnitude of S will depend upon the
selected q l u e of A. Thus, we can say that S is a fhction of A. We want to find
that value ofA for which S is minimum.Using calculus, this value is given by the

equation dS d2s >o.

= 0 such that -
dA dA2

(Remember that the value of a fhction is minim& when lirst derivative is zero
and second derivative is positive.)

Differentiating S with respect to A and equating to zero, we get

This implies that

d'
S
Further, it can be shown that -> 0 when A = jf.
d~~

4.2.2 Median
Median of a distribution locates a central point which divides a distribution into
two equal halves, i.e., it is the middle most value among a set of observations.
Let us start with examples in a discrete case. Consider a data set having 5 distinct
observations: 2,4,9,12, 19 (arranged in ascending order). Here 9 is the middle
most value since an equal number of observations are to its left and to its right.
I
Thus, 9 is the median of the above observations. Consider another data set having
t
6 distinct observations: 3,8,15,25,35,43. Here any point between 15 and 25
has the property that equal number of observations are to its left and to its right.
I Any point in the interval 15 to 25'may be used as a median. Conventionally we
i take the middle point of such an interval to define median uniquely. Thus 20 is
the median of 3, 8, 15, 25, 35, 43.
I
When a data set has non-distinct observations - a situation more common in
practice - difficulties may arise. In such situations, it may not be always
possible to locate the middle most value or the central point that divides the
dfstribution into two equal halves, For example, in the case of the data set having
Summarisation o f 5 observations 2, 9, 9, 12, 19 the value 9 is repeated twice. Thus, a formal
Univariate Data
definition of median is needed to overcome such difficulties.

A median of a distribution is a point or a central value such that at least

50% of the observations are less than or equal to it and at least 50% of the
observations are greater than or equal to it. With this definition of median
and the convention of taking the middle point of a class in which each point i
a median, median of a distribution can always be specified uniquely Thus, median
of observations 2,9,9, 12,19 is 9 because 3 of the 5 observations (60%) are
less than or equal to 9 and 4 of the 5 observations (80%) are greater than or
equal to 9.

Let us find out the median household size from the frequencydistribution in Table
4.1. We notice that 77 (out of 100) households have family size of less than or
equal to 4 and 56 households have family size of more than or equal to 4. Thus
median family size in this case is 4.

M d a n for a grouped Erequency distribution of a continuous variable is easier to

underbtand if one looks at the associated histogram with height of a rectangle equal
to the frequency density, ,;f of the class. In such a lustogram, the area of a
rectangle gives the hquency of the corresponding class. The median, in this case,
is a point in one of the classes such that the areas to its left and to its right are
50% each. First step is to locate the class, up to the right boundary of which the
total area is at least 50% (called the median class). Then the median is computed
by adding, to the lower boundary value of this class, the length of a part of this
class interval in proportion to the fkquency needed to achieve 50%. A convenient
method of finding out the median class is to compute the cumulative Erequency
(discussed in Unit 2, Section 2.3.3) and identifjmg the class interval in which the
N
-th observation lies.
2

This method of computing median is illustratedthrough the data on monthly average

household expenditure on food given in Table 4.2.

Histogram
1.25

--
-
.5
m
C
1.00

6 0.75 --
6
C
e,
0.50 --
2
L4
0.25 --

262.5 286.5. 310.5 334.5 r358.5 382.5

Median = 350.8
406.5 '
Monthly Household Expenditure (Rs.)

Fig. 4.1
Area up to the class boundary 334.5 is 3 1 and upto 358.5 is 59. Hence the median Measures of Central
Tendency
lies in the class 334.5 - 358.5. We now want to find a poht in this class so that
the area from 334.5 to the point is (50 - 31) = 19, where area up to 334.5 is
31. Since the rectangle over the interval 334.5 - 358.5 has an area of 28, and
19
is of length 24, to get an area of 19 we need 28th part of 24. This works out
19
to be 28 X 24 = 16.3. Thus the median is 334.5 + 16.3 = 350.8. Note also
that the area in the class 350.8 to 358.5 is 28 - 19 = 9 and to the right of 350.8
is 9 + 41 = 50, as it should be.

Based on the above procedure, we can write a formula for the computation of
median.

lmis the lower limit of the median class, i.e., the class in which median lies,
N is the total frequency,
C is the cumulative frequency of classes preceding the median class (note that
C = 3 1 in the above example),
fm is the frequency of median class, and
h is the width of median class.
4.2.3 Mode
As has been pointed out earlier, often observations tend to cluster around a central
value. A simple measure of h s phenomenon is called mode.

Mode or modal value of a discrete variable is defined as that value of the

variable for which fi-equency is maximum. Mode, however, is not the majority, i.e.,
it does not imply that most (50% or more) of the observations have the modal
value.

From Table 4.1 we find that the mode or modal value of household size is 4 as
this value occurs with largest frequency of 33 among 100 households.

There are, however, data sets when mode cannot be defined uniquely, i.e., the
distribution has multiple mode. Raw data with 7 hypothetical observations with
values 4,3,4, 1,2,5,3, have two modes, 3 and 4. Distributions having two modes
are called bimodal distributions,though the frequently encountered distributions
have only one mode or are unimodal.

For observations on the continuous variable, like monthly household expenditure

on food, no two observations are likely to have same value and so mode is not
a meaningful measure of such raw data. However, central tendency comes out
clearly when these raw data are grouped into various class intervals. For grouped
data modal class is defined as the class having largest frequency. Since large class
intervals are likely to include large number of observationsand smaller class intervals
are likely to have few observations, definition of modal class is meaningful only
when class intervals have equal length.
Summarlsation o f For discrete data it is easier to imd out the mode. But in the case of continuous
Univariate Data
data computation of the mode is done by the following formula:

lmis the lower limit of the modal class, i.e., the class in which mode lies,
A,(= f, -f,-,) is the difference of the frequencies of the modal class
and its preceding class,
A,(= fm - fm+,) is the difference of the frequencies of the modal class
and its following class, and
h is the width of the modal class.

Let us look back to Table 4.2. Here modal class is 334.5 - 358.5 as it has the
highest frequency, 28.

Thus, lm= 334.5, A, = 28 - 16 = 12,A, = 28 - 26 = 2 and h = 24.

Mode is a usefhl measure of central tendency when a frequency distribution

has a strong peak and it is particularly useless when a frequency distribution is
almost flat.
Check Your Progress 1

1) The kquency distribution of a family size for 250 families in a ward of an

industrial town is given below:

I Total 250 1
Find the mean, median and mode.
 Measures 01 Central
2) Compute the mean, median and mode for the following ihquency distribution. Tendency

Frequency Distribution of IQ for 309 Six-Year old Children

I.Q. Frequency
160 - 169 2
150 - 159 3
140 - 149 7
130 - f39
- 19
120 - 129 37
110 - 119 79
100 - 109 69
90 - 99 65
80 - 89 17
-
70 79 5
-
60 69 3
50 - 59 2
40 49
'- 1
Total 309

4.3 OTHER MEASURES OF CENTRAL TENDENCY

Besides the arithmetic mean, median and mode there are other averages which
are relatively unimportant but may be appropriate in particular situations.These
are Geometric Mean and Harmonic Mean. We will discuss these in Section 4.3.1.

Often we see that all the observations do not have equal importance. In such cases
we need to give differential importance to different items. Here we use weighted
means -arithmetic, geometric or harmonic -instead of simple means. This we
will discuss in Section 4.3.2.
4.3.1 Geometric Mean and Harmonic Mean
Often we have to deal with data that are time dependent, i.e., time series data
which are unlike one-time data of Tables 4.1 and 4.2. For time dependent data,
it is often of interest to fkd the pattern of change over time. Consider the following
two data sets.
Summarisation of
Univariate Data

First set looks like basic salary (in Rs.)of an employee for 7 years with annual
increment of Rs. 100 per year.

Second set looks more llke his gross salary (in Rs.).Annual increase in the two
sets are given below.

Set1: 100 100 100 100 100 100

Set 11: 110 121 133 147 161 177

Arithmetic mean of annual increase is 100 for Set I and 141.5 for Set II. On the
basis of these average annual increases, if one works-out figures for the two sets,
starting fiom the initial values, one would get the following.

S e t I : 1000 1100 1200 1300 1400 1.500 1600

Set II : 1100 1241.5 1383 1524.5 1666 1807.5 1949

That the use of arithmetic mean has worked well for Set I and not for Set 11is
because the progression of original numbers in the two sets are different. In Set
I, increment has been a fixed quantum whereas in Set 11, figures have increased
at a fixed rate. Fixed quantum of increase is called arithmetic progression and
arithmetic mean is appropriate to describe the increase. Fixed rate of increase is
called geometric progression and geometric mean is most appropriate to describe
the increase.

For n numbers X,, X,, ...Xn geometric mean (GM) is defined as the nth root
of the product of these n numbers, i.e.,

Clearly, GM is not defined unless all the n numbers are positive. By taking logarithm
of GM, one has

whkh shows now GM can be computed by using a log-table. Anti-logarithm of

the arithmetic mean of log Xvalues is GM. For the second data set, gross salary
incr'eased at the rate of 11% every year. In practice, however, increaseldecrease
will not be at a fixed rate over the years; and it is meaningful to talk about average
rate because fixed rate situation is rare. In general, GM is more appropriate average
for percentage (or proportionate) rates of change than arithmetic mean as in the
case of rise in various price indices, cost of living indices, etc.

F d y , we discuss about another measure of location called harmonic mean 0.

This mean comes naturally in many situations as in the following illustration.A stockist
stoaks Rs. 5000 worth of an item at the beginning of every month. Unit rate
(in Rs.)of the item for five successive months had been 10.75, 11.SO, 14.00,11.45
and 12.00. The stockist wants to find average rate per unit of the item he has Measures of Central
Tendency
stocked for five months. Computation is presented below :

Month Amount Spent Unit Rate

(Rs.1 (Rs.1
1 5000 10.75
2 5900 11.80
3 5000 14.00
4 5000 11.45
5 5000 12.00
TOG!! 25000

Total Money Spent

Average price (in Rs.) of his entire stock = Total Quantity

The last expression is the reciprocal of the arithmetic mean of reciprocals and
is called harmonic mean (HM). For a set of n values X,, X,,.....Xu, HM is
defined as

Note that HM is not defined when any observation is zero.

If the stockist, instead of stocking Rs.5000 worth of items, stocks 3000 items
at the beginning of every month at the given prices, the appropriate average would
be arithmetic mean. To verifL this, we can write

Total Money Spent

Average Price = Total Quantity Purchased

- 10.75+ 11.80+ 14.00+ 11.45+ 1200

= AM of the given prices.
5

4.3.2 Weighted Means

For many practical applications weighted means (arithmetic, geometric or harmonic)
reflect hen omen on more clearlv than unweihted or sirn~lemeans that have been
Summarisation o f computed so far. For computation of, say, consumer price index, not all
Univariate Data
commodities are equally important. I n m e in f k l cost may affect consumer price
index more than an i n m e in agriculturalprices. For stock market, stock of some
key companies may be a trend setter. Weighted means are more appropriate in
such situations. To find weighted mean, a weight wi is attached to each XI: and
the means are computed as if wiYsare,~symbolically,frequencies of the
correspondingq s . The computational formulae are as given below:

i=l
Weighted AM =
e
i=l
wi

weighted GM = (fi XP')=

i-1 and

i=l
Weighted HM = T L
w.

Weighted mean becomes equal to unweighted mean when each wi is same or

equal to unity.
4.3.3 Pooled Mean
O f t h we come across situations when means have been computed for different
sources or samples. In such situations we become interested to find an overall
mean if it is meaningful. This is done by computing what is called apooled mean.
The procedure of computing a pooled mean is given below.

Let m,, m,, ..... mr be r arithmetic (or geometric or harmonic) means, computed
on the basis of n,, n,, .....nr observations respectively. Then

r
1
Pooled arithmetic mean = ; Cmini, where n = ni
i-1 i-1

-1
Pooled geometric mean = (Qm y ) and

Pooled harmonic mean = 3

i-1 mi

where n = n, + n, + ..+ nr
Note that the above expressions are similar to the expressions for weighted
me2ins.
 bf Central
4.3.4 Choosing a Measure of Central Tendency Measures c
Tendency

It has already been discussed when a particular mean, AM or GM or HM, is

more appropriate than the other two. However, when one has grouped data in
which either of the end classes are open ended, i.e., of the type 'upto c,' and /
or 'ck-,and above', mid-points of such classes cannot be computed. Consequently,
no mean can be computed. There is, however, no problem in computing median
or mode in such cases. On the other hand, a pooled median or mode cannot be
computed, like the case for mean, unless all the sdts of data are made available
in their entirety. These problems are related to compu't&onal difficulties and not
to appropriateness of measure.

Since graphical representation of data is more appealing, median or mode are more
usehl in such a situation because their crude values can be obtained easi\ywithout
having to go through any computations.Also, median and mode are simple concepts
for communication and comparison between griiphs. It has, however, been
observed that median is less stable than arithmetic mean in repeated sampling and
one needs to be careful when comparing graphs.

For data that has a distribution close in shape to what is called normal with one
peak and going down symmetrically on either side, one may use one of mean,
median or mode because for a normal shape distribution, these measures have the
same value.

It should be clearly understood that choosing an appropriate measure of central

tendency is not an end to data analysis, and much still remains. For example, by
saying that household average monthly expenditure on food is Rs. 348.66, it does
not say whether a large number of households have very low monthly average
expenditure on food or a few households have a very good menu. Next set of
analysis aims at answering such questions.

4.4 PERCENTILES
Concept of percentiles will be explained by using mainly Table 4.2 data on average
monthly household expenditure. Percentiles are used in two directions, depending
on the question to be answered. Direction of a question may be, what per cent
of households have monthly average food expenditure upto Rs.350.80? Or it may
be, what is the maximum monthly average food expenditure of the lower 50% of
the households? Note, from our earlier computation of median of Table 4.2
distribution, that the answer to one question is the figure in the other,.i.e., 50%
of the households have Rs.350.80 as maximum average monthly food expenditure.
Depending on interest, percentage below a cut-off point may be called for : when
a poverty line is decided, it is of interest to h o w the percentage below the poverty
line. In the other direction, it may also be of interest to find the status of lower
10% or upper 5% of the population. These are answered by using what are called
percentiles.
4.4.1 Percentile: Definition and Computation
For any given percentage v, vth percentile is PV,a value of the variable being
stud:ed, so that at least vpercent of the observations are less than or equal to
PVand at least (100 - v) percent of the observations are greater than or equal
to P .
Summarisation of For example, for Table 4.1, distribution of household size, Py= 5 for any v
r
Univariate Data
from 78 to 89. I
4
For grouped data, percentiles are more clearly understood when one looks at the
cumulative distribution hction. Let F(X) be the proportion of observations less
than or equal to X, Any given value Xois then the 100 F(XJth percentile. For
Table 4.2, class boundaries, one has F(286.5) = 0.01, F(3 10.5) = 0.15, F(334.5)
= 0.31, F(358.5 ) = 0.59 and F(382.5) = 0.85, and consequently Rs. 286.5 =
PI,, Rs.310.5 = P,,, Rs. 334.5 = P,,, Rs. 358.5 = P,,, and Rs. 382.5 = P,,.
Note that any amount less than Rs.262.5 (lower boundary of first class interval)
is zero-th percentile and amount more than Rs.406.5 (upper boundary of last
class interval) is 100th 3ercentile.
4.4.2 Quartiles and Deciles
Depending on its use, some specific percentiles go by different names. Evary 25th
percentile is called a quartile, and every 10th percentile is called a decile. For
example,

25th percentile = P,, = Q, = first quartile

50th percentile = P,, = Q2= second quartile
75th percentile = P,, = Q3 = third quartile
10th percentile = PI, = dl = first decile
20th percentile = P2,= d2 = second decile, etc., and
P50 = Q2 = d, = median.

The formulae for Q, and Q3a& similar to the formula for the median. ~ h e s e
can be directly written as given below.

4
Q3 = l ~ , -
+
xh,
fa,

where C denotes the cumulative frequency of classes preceding the fixst (or third)
quartile class and h is the corresponding class width.

Using similar notations, it is possible to write the formula for any partition value.
For example, the formula for 40th percentile can be written as

Percentiles also go by the name of fractiles when proportions, instead of .

percentages, are used. For example, P3,is 0.3 fiactile.

Just as one does not get a complete picture of a distributionby looking at a measure
of location, too many percentiles may be needed to describe the spread or
dispersion of a-distribution. It is felt that there should be some simple measures
of dispersion. This is the topic of discussion of the next unit.
Check Your progress' 2 Measures of Central
Tendency

1) Given-below are the prices in ratios for five commodities with the
corresponding weights. Calculate the Weighted Arithmetic. Mean and
Geometric Mean.

Commodity Price Ratio Weight

1 2.20 30
2 1.85 25
3 1.80 22
4 2.05 13
5 1.75 10

2) The earnings of five nationalised banks, in crores of rupees, is given below.

217.40 330.50 682.55 1263.59 2249.63

Find the Geometric Mean of the earnings.

...................................................................................................................

3) The distribution of age of males at the time of marriage was as follows :

Age (years) No. of Males

18 - 20 5
20 - 22 18
22 - 24 28
24 - 26 37
26 - 28 24
28 - 30 22

Find at the time of maniage (i) the average age, (ii) modal age, (iii) the median
age, (iv) third quartile, (v) sixth decile, (vi) nineteenth percentilz.
Summarisation of
Univariate Data
4) In a factory, a mechanic takes 15 days to fabricate a machine, the second
mechanic takes 18 days, the third mechanic takes 30 days and the fourth
mechanic takes 90 days. Find the average number of days taken by
the workers to fabricate the machine. Which average would you use, and
whfl

5) The amount of interest paid on each of the three different sums of money
yielding lo%, 12% and 15% simple interest per annum are equal. What
is the average yield percent on the total sum invested?

4.5 LET US SUM UP

In this unit you have learned to compute various measures of central tendency.
These measures of central tendency can be divided into two broad categories,
namely mathematical averages and positional averages. Positional averages ike
mode, median, quartiles, percentiles, etc., while arithmetic mean, geometric mean
and harmonic mean are mathernatid averages. Geometric Mean is most suitable
for averaging ratio and proportional rates of growth while Arithmetic mean or
Harmonic mean can be used to find average rates like price, speed, etc. depending.
upon the nature of the given condition.

4.6 KEYWORDS
Arithmetic Mean :Sum of observed values of a set divided by the number of
observations in the set is called a mean or an average.

Frequency Distribution :The arrangement of data in the form of frequency

distribution that describes the basic pattern which the data assumes in the mass.

Geometric Mean :It is the mean of n values of a variable computed as the nth
root of their product.

Harmonic Mean :It is the inverse of the arithmetic mean of the reciprocals of
*

the observations of a set.

 --
Median. : In a set of observations, it is the value of the middlemost item when Measures of Central
Tendency
they are arranged in order of magnitude.

Mode : In a set of observations, it is the value which occurs with maximum

fiequenc y.

4.7 SOME USEFUL BOOKS

Elhance, D. N. and V. Elhance, 1988, Fundamentals of Statistics, Kitab Mahal,
Allahabad.
Nagar, A. L. and R. K. Dass, 1983, Basic Statistics, Oxford University Press,
Delhi
Mansfield, E., 1991, Statistics for Business and Economics: Methods and
Applications, W.W. Norton and Co.
Yule, G U. and M. G Kendall, 1991,An Introduction to the Theory of Statistics,
Universal Books, Delhi.

4.8 ANSWERS OR HINTS TO CHECK YOUR

PROGRESS EXERCISES
'&heckYour Progress 1

Check Your Progress 2

1) Rs. 1.96 ; Rs. 1.95

2) Rs. 674.31 crores /

3) (i) 25.83 years (ii) 24.82 years (iii) 24.86 years (iv) 27.30 years
(v) 25.59 years (vi) 28.79 years
4) Arithmetic Mean, 38.25 days
5) Harmonic Mean, 12%.