MEASURES OF CENTRAL TENDENCY

Arithmetic Mean

x1  x2  x3  ...  xn  x
Discrete Observations x 
n n
 fx
x
N
Frequency Distribution In case of simple frequency x = individual values
distribution
In case of grouped frequency x = mid-point of class intervals
distribution
N = total number of observations
N  f
 fd
x  A C
N
Assumed Mean / Step- A = Assumed Mean Take any value of x as A – preferable any
Deviation Method C = Class Length value in the middle
d = deviation
x A
d
C
Property 1 If all the observations are constant, AM is also constant
the algebraic sum of deviations of a set of observations from their AM is
Property 2
zero
AM is affected both due to change of origin and scale
Property 3
If 𝑦 = 𝑎 + 𝑏𝑥 then 𝑦 = 𝑎 + 𝑏𝑥
Combined AM
n1 x1  n2 x2
Property 4
xc 
n1  n2
 AM is best measure of central tendency
 AM is based on all observations
General Review  AM is affected by sampling fluctuations
 AM is amenable to mathematical property
 AM cannot be used in case of open end classification
 Median

 If n = odd, then middle term

Discrete Observations
 If n = even, average of two middle terms
Simple Frequency  First make column of less than cumulative frequency
Distribution  Apply same formula as discrete
 Step 1: Prepare a less than type cumulative frequency
distribution.
 Step 2: Calculate N/2 and check between which class boundaries
it falls and call it as Median Class.
 Step 3: Obtain Below values from Median Class
𝒍𝟏 𝑵𝒖 𝑵𝒍 C
LCB of Cum Freq. of Cum. Freq. of Class length
Median Class Median Class Pre-Median of Median
Grouped Frequency Class Class
Distribution
 Step 4: Apply main formula
 N 
 2  Nl 
Me  l1   C
 Nu  Nl 
 
For a set of observations, the sum of absolute deviations is minimum,
when the deviations are taken from the median.
Property 1
 xi  Me
Property 2 Median is also affected by both change of origin and scale.
 Median is also called as positional average
 Median is not based on all observations
General Review  Median is not affected by sampling fluctuations
 Median is best measure of central tendency in case of open end
classification
 Partition Values

 These may be defined as values dividing a given set of observations into

number of equal parts
 When we want to divide the given set of observations into two equal parts,
we consider median, similarly there are quartiles, deciles, percentiles

Name of PV No. of equal No. of PVs Symbol

Meaning parts
Median 2 1 Me
Quartile 4 3 Q1 , Q2 , Q3
Decile 10 9 D1 , D2 ,..., D9
Percentile 100 99 P1 , P2 ,..., P99
 Main –
(𝒏 + 𝟏)𝒑𝒕𝒉 term

Formula –  Value of p depends on partition value

Discrete # Median Quartile Decile Percentile
Observations First 1/2 1/4 1/10 1/100
Second 2/4 2/10 2/100
…
Last 3/4 9/10 99/100
  Formula is like Median with small change
 Finding Relevant Class like Median Class
Median
 N 
 2  N l 
Me  l1   C
N
 u  N l 
 
Quartile 1
 N 
  N l 
Q1  l1   4 C
 Nu  Nl 
Formula –
Grouped
Frequency  
Distribution Quartile 3
 3N 
 4  N l 
Q3  l1   C
 uN  N l 
 
Decile 9
 9N 
  N l 
D9  l1   10 C
 Nu  Nl 
 

Mode

Meaning Mode is the value that occurs the maximum number of times
 If two or more observations are having maximum frequency
then there are multiple modes [multimodal distribution]
 If there are exactly two modes then distribution is called as
Special Thing about Mode Bimodal Distribution
 If all observations are having same frequency then distribution
has no mode
 We can say that Mode is not rigidly defined
 Find Modal Class: Class with highest frequency and obtain below
values
Grouped Frequency f f f
Distribution frequency of pre frequency of the frequency of the
modal class modal class post modal class
 Apply Formula
  f 0  f 1 
Mo  l1   C
2
 0 f  f 1  f1 

Property 1 If all the observations are constant, mode is also constant

Property 2 Mode is also affected both due to change of origin and scale
 Mode is not based on all observations
General Review  Mode is not rigidly defined
 Mode is not amenable to Mathematical Property

Relationship between Mean, Median and Mode

In case of Symmetric
Mean = Median = Mode
Distribution
In case of Moderately
Skewed Distribution Mean – Mode = 3 (Mean – Median)
(Empirical relationship)

Geometric Mean

For a given set of 𝑛 positive observations, the geometric mean is

Definition
defined as the 𝑛 root of the product of the observations

G   x1  x2  ...  xn 
1/ n
Formula – Discrete
1

 
Formula – Frequency
Distribution G  x1  x2  ...  xn
f1 f2 fn N

Logarithm of G for a set of observations is the AM of the logarithm of the

observations
Property 1 1
log G   log x
n
Property 2 If all the observations are constant, GM is also constant
Property 3 GM of 𝑧 = GM of 𝑥 × GM of 𝑦
GM of
Property 4 GM of 𝑧 =
GM of
 Harmonic Mean

For a given set of non-zero observations, harmonic mean is defined as

Definition
the reciprocal of the AM of the reciprocals of the observation
n
H
Formula – Discrete 1
( )
x
N
Formula – Frequency H
f
Distribution
( )
x
Property 1 If all observations are constant HM is also constant
Combined HM
n1  n2
Property 2 xc 
n1 n2

H1 H 2

Use of GM and HM

Both Both are used for calculating average rates

GM Appropriate for rates having percentages
HM Appropriate for rates other than percentages

MEASURES OF DISPERSION

 Dispersion for a given set of observations may be defined as

Meaning of Measure of
 the amount of deviation of the observations,
Dispersion
 usually, from an appropriate measure of central tendency
Absolute  These are with units
Measures of  These are not useful for comparison of two
Dispersion variables with different units.
 Example: Range, Mean Deviation, Standard
Deviation, Quartile Deviation
Types of Measure of
Relative  These are unit free measures
Dispersion
Measures of  These are useful for comparison of two
Dispersion variables with different units.
 Example: Coefficient of Range, Coefficient of
Mean Deviation, Coefficient of variation,
Coefficient of Quartile Deviation
 Range

LS
Discrete – Formula  L: Largest Observation
 S: Smallest Observation

Grouped Frequency LS

Distribution – Formula  Largest Observation = UCB of last class interval
 Smallest Observation = LCB of first class interval
LS
Coefficient of Range  100
LS
 Not affected by change of origin
 Affected by change of scale (only value)
Property 1
 No impact of sign of change of scale
 Note: Measure of Dispersion can never be negative
 Not Based on All Observations
General Review
 Easy to Compute

Mean Deviation

 Mean deviation is defined as the

 arithmetic mean of the
Meaning
 absolute deviations of the observations
 from an appropriate measure of central tendency
1
Formula – Discrete MDA   | x  A |
n
 A = Appropriate Central Tendency Measure

Formula – Frequency 1
Distribution MDA   f | x  A|
N
Coefficient of Mean Mean Deviation about A
Coefficient of Mean Deviation = × 100
Deviation A
Mean Deviation takes its minimum value when deviations are taken
Property 1
from Median
Property 2 Change of Origin – No Affect, Change of Scale – Affect of value not sign
 Based on all observations
 Improvement over Range
General Review  Difficult to compute
 Not amenable to Mathematical Property because of usage of
Modulus
 Standard Deviation

 Improvement over Mean Deviation

Meaning  It is defined as the root mean square deviation when the
deviations are taken from the AM of the observations

( x  x ) 2
 x  SDx 
Formula – Discrete
n
 x2
x
2
 x  SDx 
n
 f ( x  x )2
 x  SDx 
Formula – Frequency N
Distribution
 fx 2
x
2
 x  SDx 
N
SDx
Coefficient of Variation  100
x
|a b|
SD for any two numbers SD 
2
SD for first n natural
n2  1
numbers s
12
Property 1 If all the observations are constant, SD is ZERO
No effect of change of origin but affected by change of scale in the
Property 2
magnitude (ignore sign)

n1s12  n2 s2 2  n1d12  n2 d 2 2
SDc 
Property 3
n1  n2
d1  xc  x1
d 2  xc  x2
 Quartile Deviation

Q3  Q1
Formula QDx 
2
Calculation Quartiles are calculated same as we studied in Central Tendency
Q3  Q1
Coefficient of Quartile
 100
Deviation
Q3  Q1
 It is the best measure of dispersion for open-end classification
 It is also less affected due to sampling fluctuations
General Review
 Like other measures of Dispersion, QD is also not affected by
change of origin but affected by scale ignoring sign