Analysing and Interpreting

Learner’s Performance UNIT 16 CORRELATION :

IMPORTANCE AND
INTERPRETATION
Structure
16.1 Introduction
16.2 Objectives
16.3 The Concept of Correlation
16.4 Co-efficient of Correlation
16.4.1 Maximum Range of Values of Co-efficient of Correlation

16.5 Types of Correlation

16.5.1 Positive and Negative Correlation
16.5.2 Simple, Multiple and Partial Correlation
16.5.3 Linear and Curvilinear Correlation

16.6 Methods of Computing Co-efficient of Correlation (Ungrouped Data)

16.6.1 Spearman’s Rank Difference Coefficient of Correlation ()
16.6.2 Pearson’s Product Moment Co-efficient of Correlation (r)
16.6.3 Pearson’s Product Moment Co-efficient of Correlation (Scattergram)

16.7 Interpretation of the Co-efficient of Correlation

16.8 Misinterpretation of the Co-efficient of Correlation
16.9 Factors Influencing the Size of the Correlation Co-efficient
16.10 Importance and Use of Correlation in Educational Measurement
and Evaluation
16.11 Let Us Sum Up
16.12 References and Suggested Readings
16.13 Answers to Check Your Progress

16.1 INTRODUCTION
In Units 13 & 14, we have discussed those statistical measures that we use
for a single variable i.e. the distribution relating to one quantitative variable. Now,
we shall study the problem of describing the degree of simultaneous variation
of two or more variables. The data in which we secure measures of one variable
for each individual is called a univariate distribution. If we have pairs of measures
on two variables of each individual, the joint presentation of the two sets of
scores is called a bivariate distribution.
You may want to study the possibility of a relationship between two variables
and see what kind of relationship exists. For example if you want to study
70 the relationship between height and weight - whether the change in one will
bring a change in other or not. Or if you want to find the relationship between Correlation: Importance
and Interpretation
hours of study and achievement, sex and enrolment etc., you can do so by
finding correlation between them. In simple words, we can say that the statistical
tool which helps us to study the relationships between two or more than two
variables is called correlation. According to Tuttle, “Correlation is an analysis
of the co-variation between two or more variables”. With the change in one
variable, the other related variable changes is known as correlation. You know
that as the child grows, the height as well as weight increases. Similarly, if a
person has more height than the other, the person is likely to have more weight
than the latter and we can say height and weight are positively correlated. You
will study about correlation in greater details in this unit.

16.2 OBJECTIVES
After reading this Unit, you will be able to:
 define correlation;
 define co-efficient of correlation;
 recognize various types of correlation;
 calculate the co-efficient of correlation according to nature of scores and
their distribution;
 interpret the results obtained i.e. interpret the Co-efficient of correlation;
 take necessary precautions in interpreting the co-efficient of correlation; and
 discuss the importance of correlation and its coefficient.

16.3 THE CONCEPT OF CORRELATION

Let us take an example of the scores of 5 students in mathematics and physics.
What pattern do you find in the data? You may notice that in general those
students who score well in mathematics also get high scores in physics. Those
who are average in mathematics get just average scores in physics and those
who are poor in mathematics get low scores in physics. In short, there is a
tendency for students to score at par on both variables. Performance on the
two variables is related; in other words the two variables are related, hence
co-vary.
If the change in one variable appears to be accompanied by a change
in the other variable, the two variables are said to be co-related and
this inter-variation is called correlation.
You must have understood from the above discussion that correlation is a
statistical technique that helps us know whether there is any relationship or not
between any two pairs of variables and how strong is this relationship. This
relationship is perfect or not can be determined by further analysis of
observations. This is a measure of simultaneous variation of variables described
by an integer. This can also be called the measure of relationship. If you take
measure of every subject (individual) of your sample on two different variables
e.g. if you take the marks of every student in Urdu and Mathematics in a Class
VI of your school and find out the relationship between these marks of Urdu
71
Analysing and Interpreting and Mathematics, relationship which is found out is called correlation. Such
Learner’s Performance
a data where we study two variables at a time is called bi-variate data. In
this kind of data scores of one individual in one subject could be paired with
the scores of the same individual in another subject. In this case scores of Urdu
can be paired with scores of Mathematics. This relationship is important as the
scores of the individual may change with the scores of other or the relationship
could be due to some common factor between them.
You will find that when one of these variables increases or decreases on
observation, the other paired variable is also changed proportionately. However,
it may not be taken as change of one variable with the manipulation of other
variable. When you are using correlation, you should not assume that a change
in one variable causes a change in the other. It is not a cause and effect
relationship. Even a high degree of correlation does not necessarily mean that
a relationship of cause and effect exists between the variables. Thus the
correlation is simply a degree of relationship between the variable of a
bi-variate data. This correlation can tell you just how much of the variation
in one variable is related to the other paired or correlated variable. Like all
statistical techniques correlation is only appropriate for certain kind of data.
Correlation works only with the quantifiable data in which numbers are
meaningful, usually in quantities of some sort. It cannot be used for purely non-
quantifiable data like categorical data such as gender, socio economic status,
goodness, etc. Correlation is just a co-variation and does not manifest any kind
of causation of functional relationship.

16.4 CO-EFFICIENT OF CORRELATION

The degree of association or the degree of relationship between two variables
is measured quantitatively in the form of an index which is termed as co-efficient
of correlation.
Co-efficient of correlation is a single number that tells us to what extent
the two variables are related and to what extent the variations in one
variable changes with the variations in the other.
This coefficient of correlation is determined to find the relationships because
it is simple to understand and convenient to express. It is a constant.
Symbol of co-efficient of correlation
The coefficient of correlation is always symbolized either by r or r (Rho). The
notion ‘r’ is known as product moment correlation co-efficient or Karl Pearson’s
Coefficient of Correlation. The symbol ‘r’ (Rho) is known as Rank Difference
Correlation Coefficient or Spearman’s Rank Correlation Coefficient. Sometimes
it is also written as rxy which means coefficient of correlation between x and
y variables.

16.4.1 Maximum Range of Values of Co-Efficient of

Correlation
The measurement of correlation between two variables results in a maximum
value that ranges from –1 to +1, through zero. The ±1 values denote perfect
coefficient of correlation.
72
 Correlation: Importance
Check Your Progress 1 and Interpretation

Note: a) Write your answers in the space given below.

b) Compare your answers with those given at the end of the Unit.
1. What do you mean by the term correlation?
....................................................................................................
....................................................................................................
....................................................................................................
2. What is Co-efficient of Correlation?
....................................................................................................
....................................................................................................
....................................................................................................
3. What can be the possible range of Co-efficient of correlation?
....................................................................................................
....................................................................................................
....................................................................................................

16.5 TYPES OF CORRELATION

Correlation can be classified in several different ways. Three most common ways
of classification can be:
(i) Positive and negative Correlation
(ii) Simple, partial and multiple Correlation
(iii) Linear and non-linear Correlation
Before we come to the classified types of correlation, many times you will come
across two more terms about correlation i.e. (i) no correlation and (ii) perfect
correlation. We will study about them first, before we come to types of
correlation.
(i) No Correlation- when there is no relationship observed or established
between two variables i.e. the change in one variable cannot establish the
change of other variable in a definite pattern. It may first time increase,
second time decrease and at third time may not change. In such cases,
we can say no definite change or Zero correlation. Most of the times one
variable changes and other remains constant. Sometimes, it may even be
a chance that there is a relationship.
(ii) Perfect Correlation – When change in one variable brings similar change
in the other variable under paired variate, it can be said to have perfect
correlation, that is, it has correlation value = one. In such cases the variable
changes in a constant proportion. Specifically, when the coefficient correlation
is ‘+1’, it is called as perfect positive correlation and when it is ‘-1’, it
is called as perfect negative correlation.
Now let us come to classified correlation. 73
Analysing and Interpreting 16.5.1 Positive and Negative Correlation
Learner’s Performance
As the name clearly suggests they are directional associations or correlations
which means the two variables of a bi-variate data vary in the same direction
or the different direction. You will understand it better with the following
explanations about them one by one.
(i) Positive Correlation: If both the paired variables under study vary in the
same direction with any change, then, they are said to be positively correlated.
It means that if one of them increases, the other related variables also increases
and if one of them decreases, the other also decreases. Both of them have
similar direction of change. For example, height is positively correlated to weight.
Similarly, height and weight are positively correlated to age. In education, the
motivation or intelligence may be positively correlated to achievement. Other
examples could be reading or study habits and achievement, attendance and
achievement, etc. This positive correlation can further be explained with the
following example.
Example:
In a class the marks of ten students were studied in Urdu and History to know
whether the marks in Urdu have some relationship with the marks obtained
in History or to know the correlation between marks of two subjects. We got
the marks of ten students as under:

Student Marks in Urdu Marks in History

1 25 35

2 20 30

3 35 45

4 28 39

5 45 50

6 36 47

7 50 65

8 18 25

9 12 20

10 40 48

Now you can see that the marks of Urdu and History achieved by all the ten
students are moving in the same direction i.e. increasing/decreasing in the same
order. Thus, we say achievement of students in Urdu and History is positively
correlated.
(ii) Negative Correlation: If the two paired variables under study vary in
opposite direction with any change, then they are said to be negatively correlated.
It means that if one of them increases, the other related variable will decrease
and if one of the related variable decreases, the other will increase. Both the
74
variables have opposite direction of change. For example, the supply and the
prices of commodities. You know, if the commodity is available in abundance,
the prices fall and if it is in shortage, the prices will rise. The same could be Correlation: Importance
and Interpretation
the case with absenteeism and achievement. More the students absent themselves
from the school, the lesser will be the achievement and vice versa. These are
common examples which must have clarified the concept of negative correlation.
You could give some more examples of negative correlation at your end. Negative
correlation is not very common in our studies but it may appear, needing
explanation in many of your studies.
16.5.2 Simple, Multiple and Partial Correlation
This classification is based on the number of variables being studied.
(i) Simple Correlation: In simple correlation only two variables are studied.
For example, in most of bi-variate studies like the correlation between height
and weight, correlation between the hours of study and achievement,
correlation between the method of teaching and achievement of students
etc., we have to find simple correlation.
(ii) Multiple Correlation: In multiple correlation three or more variables are
studied simultaneously. For example, when you want to study method of
teaching, sex, social-economic status and achievement and find the
correlations among them simultaneously, you have to see the effect of
interactions among the variables along with correlation and these are studied
through multiple correlations.
(iii) Partial Correlation: Partial correlation is also used in study of correlation
of more than two variables, but in this case it is supposed that we want
to study only two variables which are said to be influencing each other,
the effect of other influencing variables being kept constant. For example,
if we want to study the effect of hours of study and achievement but the
third variable intelligence is all the time influencing, then its effect has to
be kept constant or partialled out.
16.5.3 Linear and Curvilinear Correlation
(i) Linear Correlation: Most of the correlations you will deal with in your
research studies will be linear correlations and in common sense correlation
is generally understood by the term simple linear correlation only. If the
amount of change in one variable tends to bear a constant ratio to the
amount of change in the other related variable, then the correlation is said
to be linear correlation.
Example:
The scores obtained by students of Class VIII on two variables say intelligence
and achievement are as under:

Intelligence (IQ) Achievement

50 25
60 30
70 35
80 40
90 45
75
Analysing and Interpreting Now it is clear from the above data that the ratio of change between one variable
Learner’s Performance
to the other is the same. If such a data of the variable quantities is plotted
on a graph, you will get a straight line.
(ii) Curvilinear Correlation (Non – Linear Correlation): Now it would
be clear that if the amount of change in one variable does not bear a
constant ratio with the amount of change in the other related variable, the
correlation is said to be curvy – linear Correlation.
Example:
The following data in hours of study and the achievement scores of five students
are given below :

Hours of Study Achievement Score

1 60

2 65

3 72

4 78

5 83

We will find a non-linear relationship between the two variables. If we plot

a graph of such a data, we would get a curve.
Normally, it is assumed that a linear relationship exists between the two variables
under study because the techniques of correlation for curvi-linear correlation
are complicated and do not bring much difference.
Check Your Progress 2
Note: a) Write your answers in the space given below.
b) Compare your answers with those given at the end of the Unit.
4. Suggest examples of the following:
a) Perfect positive correlation.
...............................................................................................
...............................................................................................
...............................................................................................
b) Zero correlation.
...............................................................................................
...............................................................................................
...............................................................................................
c) Curvilinear correlation.
...............................................................................................
...............................................................................................
...............................................................................................
76
 Correlation: Importance
16.6 METHODS OF COMPUTING CO- and Interpretation

EFFICIENT OF CORRELATION
(UNGROUPED DATA)
In case of ungrouped data of bivariate distribution, the following three methods
are used to compute the value of co-efficient of correlation.
1. Rank Difference Co-efficient of Correlation or Spearman’s Rank Order
Co-efficient of Correlation.
2. Pearson’s Product Moment Co-efficient of Correlation.
3. Pearson’s Product Moment Co-efficient of Correlation from Scatter
Diagram.

16.6.1 Spearman’s Rank Difference Coefficient of

Correlation ()
Rank Difference Coefficient of Correlation is used only when the data is in
ordinal scale and it is calculated with the help of ranks. It may be defined as
the correlation between the ranks assigned to the individuals in two characters.
This is the most widely used because of its easiness to compute. This is denoted
by the symbol rho () and is computed as follows:
1. First of all the data being discrete is converted into Rank form. For ranking
the data, we have to start ranking from the highest to the lowest. Highest
score is give rank value as 1 and the next is given the rank 2 and so
on. We begin with giving ranks to first set of scores followed by giving
ranks to the next set of scores separately. Suppose there are two individuals
getting the same score, then their positional ranks are added and an average
rank is given to each individual. For example, if 6th and 7th highest ranked
individuals are getting the same score, then their ranks are added and
average is taken which is 6+7/2 = 6.5. So each of them is given a rank
of 6.5. Similarly, if three individuals get the same score, then their positional
ranks are added and average is taken. After completing the ranking of
first set of data, the second set of data is also awarded ranks in the same
manner. Although too many such ties will affect the size of correlation
coefficient; but usually there are not so many ties to justify the formula
that is available for correcting these ties.

2. After giving ranks to both the set of variables, find the difference between
these ranks of two set of scores. This is calculated as an absolute score
because the sign of these differences does not effect and is of no importance
as the differences have to be squared in the next step.

3. Square each of these differences in the next column.

4. Add all these difference squares.

5. Find the Rank Order Coefficient of Correlation by using the following

formula.
77
Analysing and Interpreting
6  D2
() = 1 
Learner’s Performance
N(N 2  1)

Where () = Rank Order Correlation Coefficient

N = Total number of set of scores in either set or Number of

pairs of scores.

D = Differences of ranks of each pair

Let us understand this with an example to illustrate it further.

Example:

Suppose you have administered a Personality test and a Mental Ability test to
some students of your school and you are interested in finding the degree of
association between personality and mental ability. Let us assume that you got
the following scores on the two tests. These scores are taken for convenience
only.

Scores of students on Personality Test and Mental Ability Test

[Link]. of Student Score on Personality Score on Mental Ability
1 60 60
2 54 68
3 53 40
4 49 52
5 49 51
6 47 38
7 46 51
8 45 32
9 45 39
10 45 41
11 43 50
12 41 48
13 39 36
14 38 48
15 32 40
16 32 46
17 30 37

To find the rank order correlation coefficient among these two variables, let
78 us find the ranks and form another table for computations.
 Table 16.1 : Table for Computation of Rank Order Coefficient of Correlation: Importance
and Interpretation
Correlation
Student Personality Mental Rank Rank D = D2
Score Ability Personality Mental R1-R2
Score (R1) Ability
(R2)
1 60 60 1 2 1 1

2 54 68 2 1 1 1

3 53 40 3 11.5 8.5 72.25

4 49 52 4.5 3 1.5 2.25

5 49 51 4.5 4.5 0 0

6 47 38 6 14 8 64

7 46 51 7 4.5 2.5 6.25

8 45 32 9 17 8 64

9 45 39 9 13 4 16

10 45 41 9 10 1 1

11 43 50 11 6 5 25

12 41 48 12 7.5 4.5 20.25

13 39 36 13 16 3 9

14 38 48 14 7.5 6.5 42.25

15 32 40 15.5 11.5 4 16

16 32 46 15.5 9 6.5 42.25

17 30 37 17 15 2 4

D2 =
386.5
Now in this Table:
(i) First find the ranks of Personality scores. In column P these ranks have
been calculated like 60 is the highest score which has been given rank
1. 54,53 being the next scores have been given ranks 2,3 respectively
whereas, 42 is repeated two times. So on averaging 4 and 5, both of
these are given the rank 4.5 each. Similarly, 32 is repeated two times and
is given the rank of 15.5 each on averaging 15 and 16 ranks. So all the
personality scores are given the ranks from 1 to 17.
(ii) Similarly, find the ranks of Mental Ability scores. In column Q, these ranks
have been calculated like, score 68 is given rank one, 60 and 52 have
been given ranks 2 and 3 respectively whereas, 51 appears two times
and has been given the ranks of 4.5 each. Similarly, 48 appears two times 79
Analysing and Interpreting and is given rank of 7.5 each being an average of 7 and 8. Similarly 40
Learner’s Performance
is also repeated and is given the rank of 11.5 each, being an average of
11 and 12. Thus, scores are awarded the ranks from 1 to 17.
(iii) Now find the difference of Ranks of each of the score pairs of Personality
and Mental ability in each row. Thus, the difference of ranks of each of
these ranks is column R = Column Q – Column P without sign.
(iv) In next column find the squares of the difference of ranks, i.e. S column
is the square of the Difference. Column S is (column R)2. For example,
in Table you have got D2 = 1,1,72.25, 2.25 etc. for the first four students.
Hence Column S = Column R2.
(v) Apply the values in the formula to find rank order coefficient of correlation:

6  D2
() = 1
N(N 2  1)
In this example
D2 = 386.50
N = 17
N2 = 289
() = 1 – 6×386.5
17 (289 – 1)
() = 1 – 2319.0
4896
= 1 – 0.47 = 0.53
16.6.2 Pearson’s Product Moment Coefficient of
Correlation (r)
Pearson’s Coefficient of correlation is determined when the data of variables
consists of scores paired in some meaningful way. There are several ways of
calculating ‘r’ in such cases.
(a) Deviation Score Method
In this method, coefficient of correlation is calculated on the basis of deviations
of the scores from their means and is defined by the following formula.

 xy
rxy = 1 
nSx Sy

Where rxy = Correlation coefficient between x and y.

x = Deviation of first variable scores from their mean (X – X )

y = Deviation of second variable scores from the mean (Y – Y )

n = Total number of pairs of scores
Sx = Standard Deviation of first variable scores
80 S y = Standard Deviation of second Variable Scores.
Example Correlation: Importance
and Interpretation

Suppose you have the scores of ten students in Physics and Mathematics as
follows and are interested in finding the coefficient of correlation between them.

Table 16.2 : Scores of students in Physics and Mathematics

Student Scores X= X2 Scores Y= Y2

in X – in Y – Y
Physics X Maths

1 37 -2.1 4.41 75 -2.6 6.76

2 41 1.9 3.61 78 0.4 0.16

3 48 8.9 79.21 88 10.4 108.16

4 32 -7.1 50.41 80 2.4 5.76

5 36 -3.1 9.61 78 0.4 0.16

6 39 -0.1 0.01 71 -6.6 43.56

7 40 0.9 0.81 75 -2.6 6.76

8 45 5.9 34.81 83 5.4 29.16

9 39 -0.1 0.01 74 -3.6 12.96

10 34 -5.1 26.01 74 -3.6 12.96

X=391 X2=208.9 Y=776 Y2=226.4

X = X /N = 39.1 Y = Y/N = 77.6

X 2
Sx = = 4.57
N

Y 2
Sy = = 4.76
N

In this Method,

i. Find means of both the scores i.e. scores of Physics and Mathematics
as calculated above are 39.1 and 77.6

ii. Then find deviation of each score from means as (x) and (y).

iii. Then find deviation squares in each case as calculated below and sum them
up. These values in this example are: X2=208.9 and X2=226.4
81
Analysing and Interpreting Table 16.3 : Table for Calculation of Product Moment Coefficient of
Learner’s Performance
correlation by Deviation Score Method
Student Phy Maths x y xy x2 y2
(x) (y) Deviation Deviation

1 37 75 -2.1 –2.6 5.46 4.41 6.76

2 41 78 1.9 +0.4 0.76 3.61 0.16

3 48 88 8.9 +10.4 92.56 79.21 108.16

4 32 80 -7.1 +2.4 -17.04 50.41 5.76

5 36 78 -3.1 +0.4 -1.24 9.61 0.16

6 39 71 -0.1 –6.6 0.66 0.01 43.56

7 40 75 0.9 –2.6 -2.34 0.81 6.76

8 45 83 5.9 +5.4 31.86 34.81 29.16

9 39 74 -0.1 –3.6 0.36 0.01 12.96

10 34 74 -5.1 –3.6 18.36 26.01 12.96

2 2
x = y = Total xy= X= Y=
391 776 129.4 208.9 226.4
M=39.1 M=77.6
iv. Then find standard deviation of both the group of scores i.e. scores of
Physics and Scores of Mathematics as calculated above are 4.57 and 4.76.
These are done after finding x2 and y2 and their sum.
v. Then find ‘x × y’ and their sum as done above, xy=129.4.
vi. Then find coefficient of correlation.

xy
rxy 
nSx Sy

= 129.4 / (10×4.57×4.76)
= 129.4 / 217.53
= 0.59
(b) Computation of ‘r’ by Raw Score Method
In the raw score method if you substitute the values of deviations in raw score
form and value of standard deviations in raw score form you will get the following
formula for the calculation of coefficient of correlation from raw scores :

NΣXY–  ΣX  ΣY
r=
 NΣX –  ΣX    NΣY –(ΣY) 
2 2 2 2

Let us take an example for formula and find out the coefficient of correlation.
In following example, raw scores and the steps for calculating ‘r’ by Raw score
82 method have been provided.
Example Correlation: Importance
and Interpretation
Table 16.4: Table for Computation of ‘r’ by Raw Score Method
S. Score Score X2 Y2 XY
No. in in
language History
(x) (y)
1 20 12 400 144 240
2 18 16 324 256 288
3 16 10 256 100 160
4 15 14 225 196 210
5 14 12 196 144 168
6 12 10 144 100 120
7 12 9 144 81 108
8 10 2 100 4 20
9 8 7 64 49 56
10 5 8 25 64 40

X=130 Y=100 X2=1878 Y2=1138 XY=1410

(i) First find sum of all raw scores of language and History X and Y
Where, X = 130 and Y = 100 as above
(ii) Then find squares of the scores of language and History X2 and Y2 which
are calculated in column no. 4 and 5.
(iii) Find sum of squares of raw scores of language and History as above
X2 = 1878 and Y2 = 1138
(iv) Then find multiple of raw scores of each pair (XY)
(v) Find sum of multiple of raw scores of each pair as above.
XY = 1410
(vi) Now find Coefficient of Correlation by substituting each value.

NΣXY–  ΣX  ΣY 
r=
 NΣX –  ΣX    NΣY –  ΣY  
2 2 2 2

10 1410  – 130 100 

r=
10 1878  – 130
30   10 1138 – 100  
2 2
  

14100 –13000
r=
18780 –16900 11380 –10000  83
Analysing and Interpreting
1100
Learner’s Performance

1880 1380 
= 1100 / 1610.7
= 0.68
(c) Standard Score or Z – Score Method
In this method, we neither use raw scores nor their deviations from the mean.
In stead we use the standard scores to find the coefficient of correlation. Thus
in this case first we transform both the variables into their standard scores and
then find coefficient of correlation by the following formula:

Σ  Zx Z y 
r=
N

The following steps are employed in these calculations:

(i) Find Mean of both the scores sets (X and Y).
(ii) Find deviation of each score from their respective mean which will help
you in finding standard deviation of both the groups of scores X and Y.
(iii) Now find Standard Deviation (Sx and Sy) of both the sets of scores.
(iv) Then transform both the group scores into their standard score
(Z scores) by the formula which you know

x y
Zx = and Zy =
Sx Sy

(v) Then find multiple of both the Z scores (Zx) (Zy) and sum it up
(vi) Find correlation coefficient by substituting the above value in the formula
Σ  Zx Z y 
r=
N

Let us illustrate it with the same example as in the previous case of scores
of language and scores of history.
Example
Table 16.5 : Table for the calculation of Coefficient of correlation
by Standard Score of Z – Score Method
[Link]. X Y X Y Zx Zy ZxZy
1 20 12 7 2 1.61 0.54 0.8694
2 18 16 5 6 1.15 1.62 1.8637
3 16 10 3 0 0.69 0.00 0
4 15 14 2 4 0.46 1.08 0.4968
5 14 12 1 2 0.23 0.54 0.1242
84
 Correlation: Importance
6 12 10 –1 0 –0.23 0.00 0 and Interpretation

7 12 9 –1 –1 –0.23 –0.27 0.0621

8 10 2 –3 –8 –0.69 –2.16 1.4904
9 8 7 –5 –3 –1.15 –0.81 0.9315
10 5 8 –8 –2 –1.84 –0.54 0.9936

x=130 y=100 Sx=4.34 =6.8317

ZxZy

X = 13 Y = 10 Sy=3.71
6.8317
r=
10
= 0.68
The above calculation is similar to the one calculated by other method. It does
not change for the same set of data. Thus we can say that the coefficient
of correlation calculated from deviation method, raw score method or the
standard scores method will remain unchanged because all these methods
are derived from the common formula of deviation in product moment coefficient
of correlation.

16.6.3 Pearson’s Product Moment Co-efficient of

Correlation (Scattergram)
Scattergram as you know is based on plotting the scores of two sets of paired
data on a graph or diagram in a table. By doing so, it gives us a visual picture
of the bi variate data formed by the pairs of related scores. For drawing scatter
gram, the unit of each variable is decided on the basis of its range and
accordingly, score is represented on a graph. The advantage of this method
is that it gives an ‘eye check’ on the number of factors that may influence the
value of coefficient, which you can use when you interpret your observations.
It can also give you a rough check of your coefficient of correlation, if you
have already calculated by any of above three techniques. For example, if you
find that scatter plot or scatter diagram corresponds to some value r = +0.72,
whereas your calculated value by other means comes out to be 0.42. Then
you would be aware of the discrepancy and can begin a recheck of your
calculations. In computer, it is very easy to plot scatter diagrams with the help
of simple statistical softwares. You may be curious to know the construction
of a scatter diagram. Let us understand the construction of a scatter diagram
in the following simple steps:
(1) Make a table of your set of paired scores.
(2) Decide the class interval for each set of scores and make a frequency
table for each set by making tallies. These class intervals may be between
10 and 20 but 10 to 15 are most convenient classes.
Now you have both the scores in class interval forms. It is not necessary
to have exactly the same number of class interval for both the variables,
nor is necessary to use the same class interval width in both variables.
This selection of width and number of classes depends upon the kind of
data or the scores therein. 85
Analysing and Interpreting (3) Next we tally the scores pair by pair so that two scores meet at a single
Learner’s Performance
point i.e. we enter our tally mark in the cell where these two class intervals
meet at one single place. Then we take the next pair and again we mark
a tally for the pair in the same way. Thus we make tally for each pair
of scores.
(4) Inspect your own tally marks once again to see that all tallies are marked
correctly.
(5) Now you add your tallies on either side of your table to get frequencies
and total of each of these sums will be the same. This is at the bottom
of the table for the sum of columns and on right hand side for the sum
of rows frequencies. These are row A and column P in your table.
(6) Now, we write moments on both sides of these sum total columns and
sum total rows. In these moments the highest is at the top of column and
designated as y and is at the right side in a row which is designated as
x. These are deviations from arbitrary reference points. So you get second
column and row of the bottom as moments i.e. Y’ and X’. These are
row B and column Q of table
(7) In the next column and next row multiply each moment by their frequencies.
So in next row you will get fx’ and in next column on right you will get
fy’. These are row C and column R of your table.
(8) Now in next column and next row multiply each of previous row and column
with moment again. So in next row you will get fx’2 and in next column
you will get fy’2. These are row D and column S of your table.
(9) In next column you find sum of last two rows C and D i.e. find f x’
and fx’2.
(10) Similarly, sum your column R and S i.e. find fx’ and fx’2.
(11) Now take the product of the deviations from the two arbitrary reference
points. Thus, it is product of moments as you have seen above that these
deviations from arbitrary reference points are called moments. Here the
arbitrary points are not the class interval but it is the mid point of that
class interval. These products of moments are calculated by multiplying the
y’ with frequency of x’ (total frequency).
(12) Now find the sum of these product moments i.e. x’ y’.
(13) Find the coefficient of correlation by the following formula.

x ' y '–  fx ' fy ' / N 

r=
fx ' –  fx ' / N fy '   fy ' / N 
2 2 2 2

Let us understand this scatter diagram with the help of an example.

Example
Suppose, you have scored 35 students of your class on the mental ability before
teaching and examining them on their achievement after six months. You got
the scores and you are interested in finding the correlation between their mental
ability scores and the achievement scores with the help of scattergram. The
scores of the students in Mental Ability Test and Achievement Test are as in
86 table 16.6.
 Table 16.6 : Scores of the students of a class on Mental Correlation: Importance
and Interpretation
Ability and Achievement
[Link]. Mental Ability Achievement
of the student Score Score
1. 80 61
2. 95 28
3. 94 74
4. 101 46
5. 105 44
6. 89 38
7. 106 72
8. 92 41
9. 105 49
10. 107 69
11. 111 82
12. 114 76
13. 83 39
14. 112 64
15. 91 77
16. 88 50
17. 105 55
18. 106 59
19. 105 86
20. 80 63
21. 85 31
22. 93 57
23. 85 70
24. 92 43
25. 90 70
26. 89 54
27. 85 51
28. 96 58
29. 85 63
30. 98 73
31. 101 71
32. 106 76
33. 112 76
34. 93 59

35 109 72 87
 88
Analysing
Performance
Analysing and

Table for Scattergram of Pearson’s Product Moment Correlation

Interpreting Learner’s
Learner’s Performance

X-Axis – Mental Ability

and Interpreting

25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 P Q R S T

f y’ f y’ f y’ 2 y ’ x’
24
114-116 24 1 1 6 6 36 24
5 20 25
111-113 5 1 20 1 25 1 3 5 15 75 50
12
108-111 121 1 4 4 16 12
9 6 0 6 9 12 18
105-107 9 1 6 1 0 2 6 1 91 12 1 18 1 8 3 24 72 30
102-104 0 2 0 0 0
99-101 2 12 3 13 2 1 2 2 1
0
96-98 010 01 2 0 0 0 0
0 3
93-95 6 16 02 3 1 4 1 4 4 3
6
90-92 12 26 61 8 18 4 2 8 16 10
87-89 3 3 9 27 18

Y-Axis  Achievement
12 1 12 623
4
84-86 20 1 20 41 4 14 12 2 1 2 4 4 16 64 8
81-83 20 1 20 1 5 5 25 20
78-80 12 26 2 6 12 72 12
A f 1 1 2 3 2 3 5 4 1 7 4 1 1 N=35
B x’ 6 5 4 3 2 1 0 1 2 3 4 5 6

C f x’ 6 5 8 9 4 3 0 4 2 21 16 5 6 fx’ = 33
2
D f x’ 2 36 25 32 27 8 3 0 4 4 63 64 25 36 x’ = 327
T x’ y’ 6 20 32 3 8 10 0 11 6 15 48 25 18 x’ y’ = 164
Let us understand the calculations in this scattergram table through the steps Correlation: Importance
and Interpretation
we had studied.

1. Paired set of scores are there in the form of mental ability scores and
achievement scores.

2. The classes of mental ability are formed from 78 to 116 with class intervals
of 4 as shown on Y axis.

3. The classes of achievement scores are formed from 25 to 90 with class

intervals of 5 each. They are shown on X-axis.

4. Now scores are tallied as per their value on both the axis as shown in
between the Table. We enter first score pair (80, 61) in their respective
column and row. Similarly next (95, 28) pair is placed as tally in the
respective position and so on.

5. Recheck that they have marked tally in correct place.

6. Add the tally on either side in their respective columns and rows to get
frequency as shown in row A and column P. These sums of all frequencies
of A and P will be same i.e. total N. This will recheck the tallies.

7. Now mark moments as the deviation from the middle class interval.
These are there in column Q and row B. The deviation of starting class
is ‘O’ and above the class, deviations are +1, +2, +3 .... and below the
class 1, 2, 3.... etc.

8. Now multiply each moment by their respective frequency on either side

i.e. columns and rows to get f y’ and f x’ i.e. column R and row C in
this table. It is product of

P × Q = R and A × B = C

9. Now again multiply them by moment on each column and in each row
to get S and D column and row respectively in this table. It is the product
of R × Q = S column and

B × C = D row. This is fx’ 2.

10. Now sum all the rows and columns to get their sum ( ) i.e. get f x’,
fx’ 2, y’ and y’ 2.

11. Now find the product of the moments, i.e., the product of the x’ and y’
in T column.

12. Add the product moments which are there in column T i.e. you get x’y’.

13. Now find the correlation ‘r’ by the following formula

x ' y '–  fx ' fy ' / N 

r=
fx ' –  fx ' / N fy '   fy ' / N 
2 2 2 2

89
Analysing and Interpreting 164  (33  6 / 35)
Learner’s Performance r=
327 - (332 /35)  409  6 2 / 35)
164  5.66
=
(327 - 31.11)  (409 - 1.03)

158.34
=
295.89 407.97

158.34
= = 0.46 (Approx)
347.44
Check Your Progress 3
Note: a) Write your answers in the space given below.
b) Compare your answers with those given at the end of the Unit

5. Compute correlation of the marks of 10 students in rank order

method.
Scores (Math) : 30, 37, 70, 45, 69. 55, 40, 45, 55, 63
Scores (English) : 33, 40, 64, 39, 52, 40, 36, 44, 52, 60
................................................................................................
................................................................................................
................................................................................................
6. Compute correlation of the marks of 10 students in Pearson’s
product moment deviation score method.
Scores (Sc.) : 35, 37, 60, 40, 62. 55, 40, 43, 55, 50
Scores (Soc. Sc.) : 33, 40, 61, 39, 52, 50, 36, 44, 50, 60
................................................................................................
................................................................................................
................................................................................................

16.7 INTERPRETATION OFTHE CO-EFFICIENT

OF CORRELATION
Merely computation of correlation does not have any meaning until and unless
we determine how large the coefficient need to be in order to be significant,
and what does correlation tell us about the data? Coefficient of correlation is
positive or negative depending upon the direction as explained earlier in positive
and negative correlations. Thus coefficient of correlation will have values between
+1 and –1. When the value of obtained correlation is closer to 1 or –1, then
the variables are more closely related and if it is close to 0, it means that there
is no relationship between these variables. The coefficient of correlation tells
us the intensity of linear relationship in quantitative terms. In real life situations
perfectly correlated variables are rare.
90
Full interpretation of correlation ‘r’ depends upon the circumstances, one of Correlation: Importance
and Interpretation
which is the size of the sample. All that can really be said is that when estimating
the value of one variable from the value of other; the higher is the value of
coefficient of correlation; the better will be the estimate. The closeness of
relationship is not proportional to the value of “r”. If the value of this “r” is
0.85, it does not necessarily indicate that a relationship is five times as close
as one having “r” 0.17. We can only say that they have different kind of
relationships. It will represent the proportion of common variation in two variables
both in strength and magnitude.
Coefficients of correlation are not even additive. An average of correlation
coefficients in a number of samples does not represent an average correlation
in all those samples. Thus we should not add the coefficients of correlation
to find correlation of two or more samples. But we should first convert them
into additive measures and find correlations after that.
The coefficient of correlation is interpreted in verbal description. The rule of
thumb for interpreting the size of a correlation coefficient is presented below:
Table 16.7: Correlation coefficient and its interpretation
Size of Correlation Interpretation

±1 Perfect Positive/negative Correlation

± .90 to ± .99 Very High Positive/Negative Correlation
± .70 to ± .90 High Positive/Negative Correlation
± .50 to ± .70 Moderate Positive/Negative Correlation
± .30 to ± .50 Low Positive/Negative Correlation
± .10 to ± .30 Very low Positive/Negative Correlation
± .00 to ± .10 Markedly Low and Negligible Positive/Negative
Correlation

16.8 MISINTERPRETATION OFTHE

COEFFICIENT OF CORRELATION
Sometimes, we misinterpret the value of coefficient of correlation and establish
the cause and effect relationship, i.e. one variable causing the variation in the
other variable. Actually we cannot interpret in this way unless we have sound
logical base. Correlation coefficient gives us, a quantitative determination of the
degree of relationship between two variables X and Y, not information as to
the nature of association between the two variables.
 The degree of relationship or association is not ordinarily interpretable in
direct proportions to the magnitude of the coefficient of correlations. For
example, with ‘r’ = 0.25 we cannot say one variable has one fourth
association with the other variable, or with “r” = 0.50 we can not interpret
that one variable has half relationship with the other variable. This cannot
be objectively considered. In general, change of 0.10 point in coefficient
may have greater consequences when applied to coefficients having a high
value than if the same is applied to lower values.
 The strength of relationship between two variables depends, among other
things, on the nature of measurement of the two variables as well as on
the kind of subjects being studied. Thus, it is not possible to speak of 91
Analysing and Interpreting the correlation between two variables without taking these factors into
Learner’s Performance
consideration.
 Sometimes, you may find a high correlation between two variables, which
may tempt you to think that they are substantially correlated. But in actual
practice they may not have any relationship. You have to logically see
whether such an association could exist or not. There must be some common
cause for the association between the two. Mere high coefficient is
insufficient to claim a relation between the two variables.
 Although a straight line graph between two variables is an indication of
a straight correlation or high correlation between the two variables, but
sometimes, it may not be so and in such cases it may be mere chance
or it is misleading.
 The correlation coefficient is affected by the range of talent (variability)
characterizing the measurements of the two variables. In general, the smaller
is the range of talent in two variables; the lower will be coefficient of
correlation, other things being equal. For example, in any school the
correlation between the aptitude test score and marks obtained by the
students may be +0.50. But if you find the coefficient of correlation of
the students having high scores, the coefficient may be much lower.
 Coefficient of correlation like any other statistics is also dependent upon
sampling fluctuations. It is also affected by sampling variations. Depending
upon the characteristics of a particular sample, the obtained coefficient may
be higher or lower than it will be in a different sample.
 Correlation works with quantifiable data in which the number is meaningful
as we have already explained above. It cannot be used for purely categorical
data such as gender, colours etc.
 Squaring coefficient of correlation makes it easier to understand and gives
the percent of the variation in one variable that is related to the other variable
and makes it easier to understand. If r = 0.5; then square becomes 0.25
and we can say there is 25% variation in one variable with respect to
the other variable under study. This squared correlation coefficient (r2) is
the proportion of variance in y variable that can be accounted for by
knowing X variable. One important property of variance is that it may be
portioned into separate additive parts. Like total correlation of a sample
variable X with variable Y where X is total human beings. So X if divided
into Male and Female, the total “r” can also be divided into two parts
r1 and r2 where r = r1 + r2.

16.9 FACTORS INFLUENCING THE SIZE OF THE

CORRELATION COEFFICIENT
It will be helpful for you to know that the following factors influence the size
of the coefficient of correlation and can some times lead to misinterpretation:
1. The size of “r” is very much dependent upon the variability of measured
values in the correlated sample. The greater the variability, the higher will
be the correlation, everything else being equal.
2. The size of “r” is altered, when an investigator selects an extreme group
92 of subjects in order to compare these groups with respect to certain
 behaviour. “r” obtained from the combined data of extreme groups would Correlation: Importance
and Interpretation
be larger than the “r” obtained from a random sample of the same group.
3. Addition or dropping the extreme cases from the group can lead to change
in the size of “r”. Addition of the extreme case may increase the size of
correlation, while dropping the extreme cases will lower the value or “r”.

16.10 IMPORTANCE AND USE OF CORRELATION

IN EDUCATIONALMEASUREMENTAND
EVALUATION
Correlation is important in many areas of measurement and evaluation on
education. It is of particular importance in the study of similarities and individual
differences. With this common property of correlation in mind, correlation is
used in the following contexts:
(i) For Determining Reliability:
Suppose there are two examiners examining the same class on some
parameters or in some examination. The comparison of two Examiners
or two observations of behaviour by the same examiner can be judged
with the correlation. Correlation tells how reliable the inter observer rating
of behaviour is. It tells us whether the two observers agree on the scores
given or whether there is agreement between the two obtained values
or scores.
It is not only the inter observer score but even the intra observations
by the same examiner for two or more times can be easily compared
with the help of correlation.
(ii) For Determining Validity of Scores in Predictions:
Like reliabilities, the correlation is useful in determining validity of two
or more sets of scores. A high correlation is an indicator of high validity.
For example, when you are interested in knowing to what extent the scores
of students on aptitude test are related to their achievement or performances.
If there is a high correlation between the two variables, then you can
rely on any one of the tests which can be a valid measure of the variable.
(iii) Study of Individual Differences:

These relationships and the differences can be studied through the

correlations of negative and positive kind. The concept of correlation is
basic to the theory and practice of trait measurement. For example, the
use of different types of tests like mental ability tests, aptitude tests, tests
of reasoning, mathematical ability test, etc. are all used based on their
property of correlation with achievement or performance, etc.

(iv) For predicting one or More Variables:

You might have estimated academic performance of the children of your

school on the basis of your knowledge of their aptitude or some other
test score. Even sometimes you might have predicted the scores of the
students in their Board examination on the basis of their scores in previous
examination or Pre-Board examination. If some one asks you the basis
of your prediction, you may answer the query on the basis of your 93
Analysing and Interpreting knowledge of correlation. Correlation is helpful for predicting or making
Learner’s Performance
estimation of one variable from our knowledge of the other variable.
(v) For Determining the Usefulness of Regression Line:
Coefficient of correlation is very useful and the most frequently used
measure for determining the usefulness of a regression line for the estimation
purposes. It helps in proper estimations and the partial correlation of
different variables that affect a particular variable.
(vi) Determining Prediction Validity of Tests & Measurement:
Predictive validity is itself a correlation between a set of test scores or
some other predictor with an external measure. This external measure is
called the criterion. For example, to validate the intelligence tests you obtain
a set of scores on a group of your school students and later find out
the grade point averages (GPA) that these students get during their next
examination. A correlation is then run between our two set of scores or
measurements.
(vii) Qualifying the Relationship Between Variables:
As correlation qualifies the relationship between the two variables, it also
tells us whether the two variables are related to each other or not. It
tells us about the kind of relationship, i.e., what kind of relationship exists
between two variables.
(viii) Quantification of Degree of Relationship:
The correlation not only qualifies for the relationship but also tells us about
the quantity of relationship. The coefficient of correlation tells us the
quantity of relationship. Square of coefficient of correlation (r2) is the
coefficient of determination which may even help in determining one
variable with the knowledge of the other related variable. Hence, coefficient
of correlation helps us in quantification of the degree of relationship
between two related variables.
(ix) Finding Error in Prediction:
Coefficient of correlation also gives a way of specifying the error involved
in prediction of one variable from the other. The square of coefficient
of correlation expresses the amount of variance commonly shared by the
two variables, offering a means whereby the variability of human behaviour
and performance can be explained.

16.11 LET US SUM UP

Measure of relationship of paired variables is quantified by a coefficient of
correlation. The correlation coefficient is a value between – 1.0 to + 1.0. It
should be noted that a high coefficient does not imply a cause-and-effect
relationship, but merely quantifies a relationship that has been logically established.
The size of the correlation coefficient is affected by the homogeneity of the
scores on the variables. If a relationship exists between two variables and that
relationship is linear, then the scores are more heterogeneous and the greater
the range of measurement, the greater is the absolute value of correlation.
The correlation coefficient is very useful in educational evaluation, standardizing
tests and in making predictions.
94
 Correlation: Importance
16.12 REFERENCES AND SUGGESTED and Interpretation

READINGS
Aggarwal, Y.P. (1990), ‘Statistical Methods – Concepts, Applications and
Computation’, Sterling Publishers Pvt. Ltd., New Delhi.
Anderson, Richard L. (1952), ‘Statistical Theory in Research’, McGraw hills
Book Co., New York.
Best John W and James V. Kohn (2004), ‘Research in Education’, Prentice
Hall of India Pvt. Ltd., New Delhi.
Downie, N.M. (1983), ‘Basic Statistical Methods’, HarperCollins College Div.,
New York.
Ferguson, G.A. (1974), ‘Statistical Analysis in Psychology and Education’,
McGraw Hill B.K. Co., New York.
Garrett. H.E. & Woodworth, R.S.(1969), ‘Statistics in Psychology and
Education’. Vakils, Feffer & Simons Pvt. Ltd., Bombay.
Guilford, J.P. & Benjamin, F.(1973), ‘Fundamental Statistics in Psychology
and Education’, McGraw Hill B.K. Co., New York.
Kerlinger, F.N.(1993), ‘Foundations of Behavioural Research: Educational
and Psychological Inquiry’. Holt, Rinchart & Winston, New York.
Loether, H.L. & McTavish, D.G. (1988), ‘Descriptive and Inferential Statistics:
An Introduction’, 3rd ed. Boston: Allyn & Bacon.
Sonedecor, G.W. & Cochran, W.G.(1967), ‘Statistical Methods’ (6th ed.),
Lowa State University Press, Ames, Lowa.

16.13 ANSWERS TO CHECK YOUR PROGRESS

1. Correlation is the association or relationship between two variables.
2. Coefficient of correlation is a number that tells us the extent to which the
two given variables are related or the change in one variable is accompanied
by a change in the other variable.
3. A coefficient of correlation can vary from a value of +1.00 to –1.00, through
zero.
4. Self exercise (a, b, and c)
5. Self exercise
6. Self exercise

95