College Name: JDT ISLAM COLLEGE OF ARTS AND SCIENCE

Programme / Course Name: [Link] Honours

Subject: Business Statistics
Semester: 4
Module: 2 – Correlation and Regression Analysis
Signature: Notes Created by Noor Mohammed.

MODULE 2: CORRELATION AND REGRESSION ANALYSIS

PART A: CORRELATION ANALYSIS

1. Introduction to Correlation
Definition:
Correlation is a statistical technique used to measure the relationship or association between
two or more variables.
It indicates how the value of one variable changes when the value of another variable changes.
Example: Relationship between Price and Demand, Height and Weight, Advertising
Expenditure and Sales.
Key Authors' Definitions:
A.M. Tuttle: "Correlation is an analysis of the covariation between two or more variables."
L.R. Connor: "If two or more quantities vary in sympathy so that movements in one tend to be
accompanied by corresponding movements in the other, then they are said to be correlated."
Significance of Correlation:
Helps in deriving laws of demand and supply.
Useful in business decision-making (e.g., estimating sales based on ad spend).
Helps in reducing the range of uncertainty in prediction.
Foundation for Regression Analysis.

2. Types of Correlation

A. Based on Direction of Change

1. Positive Correlation:
Both variables move in the same direction.
If one variable increases, the other also increases.
If one variable decreases, the other also decreases.
 Examples:
Income and Expenditure.
Height and Weight.
Temperature and Sale of Ice-cream.
2. Negative Correlation:
Variables move in opposite directions.
If one variable increases, the other decreases.
Examples:
Price and Demand.
Volume and Pressure of a perfect gas.
Production and Price per unit.
3. Zero (No) Correlation:
No relationship exists between the two variables.
Change in one variable does not affect the other.
Examples:
Shoe size and Intelligence.
Weight of a person and their income.

B. Based on Number of Variables

1. Simple Correlation:
Study of relationship between only two variables.
Example: Yield of wheat and amount of rainfall.
2. Multiple Correlation:
Study of relationship between three or more variables simultaneously.
Example: Yield of wheat dependent on rainfall, quality of seeds, and fertilizer used.
3. Partial Correlation:
Study of relationship between two variables while keeping other influencing variables constant.
Example: Relationship between Yield and Rainfall, keeping "Fertilizer" constant.

C. Based on Linearity
1. Linear Correlation:
The ratio of change between two variables remains constant.
Graphically, it forms a straight line.
Equation: Y = a + bX .
2. Non-Linear (Curvilinear) Correlation:
The ratio of change between variables is not constant.
Graphically, it forms a curve.

3. Methods of Studying Correlation

Scatter Diagram Method (Graphic method).
Karl Pearson’s Coefficient of Correlation (Mathematical method).
Spearman’s Rank Correlation Coefficient (Mathematical method).
Concurrent Deviation Method (Mathematical method).
(Note: Syllabus focuses on Karl Pearson and Rank Correlation).

4. Karl Pearson’s Coefficient of Correlation (r)

Also known as Pearson’s Product Moment Correlation Coefficient.
Most widely used mathematical method.
Denoted by r.

Assumptions:
Relationships between variables are linear.
Variables are affected by a large number of independent causes.
A cause-and-effect relationship exists.
Data is normally distributed.

Properties of Coefficient (r):

Range: The value of r always lies between -1 and +1.
r = +1: Perfect Positive Correlation.
r = −1: Perfect Negative Correlation.
r = 0: No Correlation.
Unit-free: It is a pure number and has no unit of measurement.
Symmetry: Correlation of X and Y is the same as Y and X (rxy = ryx ).
​ ​

Independent of Origin and Scale: Changing the origin (adding/subtracting) or scale

(multiplying/dividing) does not change the value of r.

Interpretation of r Value:

Value of r Degree of Correlation

 +1 Perfect Positive

-1 Perfect Negative

+0.75 to +1 High Degree Positive

-0.75 to -1 High Degree Negative

+0.25 to +0.75 Moderate Degree Positive

-0.25 to -0.75 Moderate Degree Negative

0 to +0.25 Low Degree Positive

0 to -0.25 Low Degree Negative

0 Absence of Correlation

5. Formulas for Karl Pearson’s Coefficient

Method 1: Actual Mean Method

(Used when arithmetic means of X and Y are whole numbers)

∑ xy
r= ​

∑ x2 ⋅ ∑ y 2 ​

Where:
ˉ
x=X −X
y = Y − Yˉ

Method 2: Assumed Mean Method (Shortcut Method)

(Used when means are fractions/decimals)

N ∑ dxdy − (∑ dx)(∑ dy)

r= ​

[N ∑ dx2 − (∑ dx)2 ][N ∑ dy 2 − (∑ dy)2 ] ​

Where:
dx = X − A (A = Assumed mean of X)
dy = Y − B (B = Assumed mean of Y)
N = Number of pairs
Method 3: Direct Method (Product Moment)
(Used when values are small)

N ∑ XY − (∑ X)(∑ Y )
r= ​

[N ∑ X 2 − (∑ X)2 ][N ∑ Y 2 − (∑ Y )2 ] ​

6. Solved Problems: Karl Pearson’s Coefficient

Problem 1: Actual Mean Method

Calculate the coefficient of correlation between X and Y.

X 2 4 6 8 10

Y 4 8 12 16 20

Solution:
ˉ ) and Mean of Y (Yˉ ).
First, calculate Mean of X (X
ˉ=
X 30
=6
5
​

Yˉ = 60
5
​
= 12
Since means are whole numbers, we use the Actual Mean Method.
Calculation Table:

X Y x=X−6 y=Y−12 x2 y2 xy

2 4 -4 -8 16 64 32

4 8 -2 -4 4 16 8

6 12 0 0 0 0 0

8 16 2 4 4 16 8

10 20 4 8 16 64 32

30 60 0 0 40 160 80

Applying Formula:

∑ xy
r= ​

∑ x2 ⋅ ∑ y 2 ​
 80
r=
40 × 160
​

80
r= ​

6400 ​

80
r= = +1
80
​

Conclusion: There is a perfect positive correlation between X and Y.

Problem 2: Assumed Mean Method (Shortcut)

Calculate Karl Pearson’s coefficient of correlation.

Price (X) 10 14 18 22 26 30

Demand (Y) 18 12 24 6 30 36

Solution:
Let Assumed Mean for X (A) = 18
Let Assumed Mean for Y (B ) = 24
N =6
Calculation Table:

X Y dx(X−18) dx2 dy(Y−24) dy2 dxdy

10 18 -8 64 -6 36 48

14 12 -4 16 -12 144 48

18 24 0 0 0 0 0

22 6 4 16 -18 324 -72

26 30 8 64 6 36 48

30 36 12 144 12 144 144

Sum 12 304 -18 684 216

Values:
∑ dx = 12
 ∑ dx2 = 304
∑ dy = −18
∑ dy 2 = 684
∑ dxdy = 216
N =6
Applying Formula:

N ∑ dxdy − (∑ dx)(∑ dy)

r= ​

[N ∑ dx2 − (∑ dx)2 ][N ∑ dy 2 − (∑ dy)2 ] ​

Numerator:

6(216) − (12)(−18)

1296 − (−216)

1296 + 216 = 1512

Denominator:

Term 1: 6(304) − (12)2 = 1824 − 144 = 1680

Term 2: 6(684) − (−18)2 = 4104 − 324 = 3780

Denominator = 1680 × 3780 = ​

6350400 = 2520
​

Final Calculation:

1512
r= = 0.6
2520
​

Conclusion: There is a moderate positive correlation (r = 0.6).

7. Spearman’s Rank Correlation (R or ρ)

Developed by Edward Spearman in 1904.
Used when variables are qualitative (e.g., beauty, honesty, intelligence) and cannot be measured
quantitatively but can be ranked.
Also used for quantitative data when ranks are preferred over values to avoid the effect of extreme
values.
Cases in Rank Correlation:
1. When Ranks are Given: Directly apply the formula.
2. When Ranks are Not Given: Assign ranks to data first.
3. When Ranks are Repeated (Tied Ranks): Use adjustment formula.

Formula (Case 1 & 2 - No Ties):

6 ∑ D2
R=1−
N (N 2 − 1)
​

Where D = R1 − R2 (Difference between ranks)

​ ​

N = Number of pairs

Formula (Case 3 - Tied Ranks):

6[∑ D2 + 1 3
12 (m1
1
− m1 ) + 12 (m32 − m2 ) + … ]
R=1−
​ ​ ​ ​ ​ ​

N (N 2 − 1)
​

Where m = Number of times a rank is repeated.

8. Solved Problems: Rank Correlation

Problem 3: Ranks Given

Two judges ranked 5 participants in a beauty contest as follows:

Judge 1 (R1​) 1 2 3 4 5

Judge 2 (R2 ) ​ 5 4 3 2 1

Solution:

R1​ R2​ D(R1​−R2​) D2

1 5 -4 16

2 4 -2 4

3 3 0 0

4 2 2 4

5 1 4 16
 N=5 Total 40

Formula:

6 ∑ D2
R=1−
N (N 2 − 1)
​

6(40)
R=1−
5(25 − 1)
​

240
R=1−
5(24)
​

240
R=1−
120
​

R = 1 − 2 = −1

Conclusion: There is a perfect negative correlation in the judgment of the two judges.

Problem 4: Ranks Not Given (Assigning Ranks)

Calculate Rank Correlation.

Marks (Stats) 40 50 60 30 80

Marks (Acc) 35 60 55 40 90

Solution:
Ranking Rule: Give Rank 1 to the highest value, Rank 2 to the second highest, and so on.

X R1​(Stats) Y R2​(Acc) D(R1​−R2​) D2

40 4 35 5 -1 1

50 3 60 2 1 1

60 2 55 3 -1 1

30 5 40 4 1 1

80 1 90 1 0 0

4
Formula:

6(4)
R=1−
5(25 − 1)
​

24
R=1−
120
​

R = 1 − 0.2 = 0.8

Conclusion: High degree positive correlation (0.8).

Problem 5: Repeated Ranks (Tied Ranks)

Calculate Rank Correlation. X: 10, 20, 20, 30, 40 Y: 15, 10, 20, 20, 20

Ranking Process:
For X:
40 (Rank 1)
30 (Rank 2)
3+4
20 appears twice (positions 3 and 4). Take average: 2
​ = 3.5. Both get 3.5.
10 (Rank 5)
For Y:
1+2+3
20 appears three times (positions 1, 2, 3). Take average: 3
​ = 2. All get Rank 2.
15 (Rank 4)
10 (Rank 5)
Table:

X R1​ Y R2​ D D2

10 5 15 4 1 1.00

20 3.5 10 5 -1.5 2.25

20 3.5 20 2 1.5 2.25

30 2 20 2 0 0.00

40 1 20 2 -1 1.00

Sum 6.50
Tie Adjustment:
1 3 6
In X: 20 is repeated 2 times (m1 = 2). Adjustment:
​

12 (2
​
− 2) = 12
​
= 0.5
1 3 24
In Y: 20 is repeated 3 times (m2 = 3). Adjustment:
​

12 (3
​
− 3) = 12
​
= 2.0
Total Adjustment = 0.5 + 2.0 = 2.5
Adjusted ∑ D2 = 6.5 + 2.5 = 9.0
Formula:

6×9
R=1−
5(24)
​

54
R=1−
120
​

R = 1 − 0.45 = 0.55

PART B: REGRESSION ANALYSIS

9. Introduction to Regression
Definition:
Regression stands for "stepping back" or "returning".
It is a statistical method used to estimate or predict the unknown value of one variable from
the known value of another variable.
Sir Francis Galton first used the term.
Utility:
Prediction (Forecasting sales, profits, etc.).
Estimation of relationship (Functional relationship).
Measure of error (Standard Error of Estimate).

Correlation vs Regression
Basis Correlation Regression

Meaning Measures degree of relationship. Predicts value of one variable based on another.

Cause & May or may not indicate cause- Clearly indicates cause and effect (Independent vs
Effect effect. Dependent).
 Limits Limited to linear relationship Wider scope for prediction.
mostly.

Coefficients One coefficient (rxy ). ​

Two coefficients (bxy , byx ).
​ ​

Origin/Scale Independent of Origin and Scale. Independent of Origin but not of Scale.

10. Regression Lines and Equations

There are usually two regression lines:
1. Regression Line of X on Y: Used to predict X when Y is known.
Equation: Xc = a + bY ​

2. Regression Line of Y on X: Used to predict Y when X is known.

Equation: Yc = a + bX ​

Why two lines?

Because the objective differs. If we minimize horizontal distances, we get X on Y. If we minimize
vertical distances, we get Y on X.
They coincide only if correlation is perfect (r = ±1).

11. Regression Coefficients (b)

byx : Regression coefficient of Y on X. It measures the change in Y for a unit change in X.
​

bxy : Regression coefficient of X on Y. It measures the change in X for a unit change in Y.

​

Properties of Regression Coefficients:

1. Same Sign: Both bxy and byx must have the same algebraic sign (+ or -).
​ ​

2. Relation to Correlation: Correlation coefficient is the geometric mean of regression coefficients.

r= bxy × byx
​ ​

(Sign of r is the same as the sign of b).

3. Magnitude: Both coefficients cannot be greater than 1 (if one is > 1, the other must be < 1).
4. Mean: The arithmetic mean of regression coefficients is greater than or equal to correlation
coefficient.
bxy +byx
≥r
​ ​

2
​

12. Determination of Simple Linear Regression Equations

We use the Method of Least Squares or mathematical formulas derived from means.

ˉ , Yˉ )
Formula Set 1: Using Actual Mean (X
(Use when means are whole numbers)
1. Regression Equation of Y on X:

(Y − Yˉ ) = byx (X − X
​
ˉ)

∑ xy
byx =
∑ x2
​ ​

2. Regression Equation of X on Y:

ˉ ) = bxy (Y − Yˉ )
(X − X ​

∑ xy
bxy =
∑ y2
​ ​

Formula Set 2: Using Assumed Mean (Shortcut)

(Use when means are decimals)
1. Coefficient byx : ​

N ∑ dxdy − (∑ dx)(∑ dy)

byx =
N ∑ dx2 − (∑ dx)2
​ ​

2. Coefficient bxy : ​

N ∑ dxdy − (∑ dx)(∑ dy)

bxy =
N ∑ dy 2 − (∑ dy)2
​ ​

13. Solved Problems: Regression

Problem 6: Forming Regression Equations (Actual Mean)

From the following data, obtain the two regression equations.

X 6 2 10 4 8

Y 9 11 5 8 7

Solution:
1. Calculate Means:
ˉ =6
∑ X = 30, N = 5, X
∑ Y = 40, N = 5, Yˉ = 8
 Means are whole numbers. Use Actual Mean Method.
Calculation Table:

X Y x(X−6) x2 y(Y−8) y2 xy

6 9 0 0 1 1 0

2 11 -4 16 3 9 -12

10 5 4 16 -3 9 -12

4 8 -2 4 0 0 0

8 7 2 4 -1 1 -2

Sum 0 40 0 20 -26

2. Calculate Coefficients:
∑ xy −26
byx =
​

∑ x2
​
= 40
​
= −0.65
∑ xy −26
bxy =
​

∑ y2
​
= 20
​
= −1.3
3. Form Equations:
Equation of Y on X:

(Y − Yˉ ) = byx (X − X
​
ˉ)

Y − 8 = −0.65(X − 6)

Y − 8 = −0.65X + 3.9

Y = −0.65X + 11.9

Equation of X on Y:

ˉ ) = bxy (Y − Yˉ )
(X − X ​

X − 6 = −1.3(Y − 8)

X − 6 = −1.3Y + 10.4
 X = −1.3Y + 16.4

Verification: r = (−0.65) × (−1.3) =

​
0.845 = −0.919 (Negative sign used).
​

Problem 7: Prediction
Find the likely sale when Advertising Expenditure is Rs. 10 Lakhs.
Correlation (r) = 0.8
Mean Sales (Yˉ ) = 100 Lakhs
ˉ ) = 8 Lakhs
Mean Adv. Exp. (X
SD of Sales (σy ) = 5
​

SD of Adv. Exp. (σx ) = 2

​

Solution:
We need to predict Sales (Y) when Adv (X) is 10.
We need the Regression Equation of Y on X.
Formula using SD:

σy
(Y − Yˉ ) = r ˉ)
(X − X
​

σx ​

Substitute Values:

5
Y − 100 = 0.8 × (10 − 8)
2
​

Y − 100 = 0.8 × 2.5 × 2

Y − 100 = 2 × 2

Y − 100 = 4

Y = 104

Answer: The likely sales is Rs. 104 Lakhs.

PART C: PRACTICE SECTION

14. Solved Examples (Mixed Problems)

Example 1: Missing Information The regression equations are 8X − 10Y + 66 = 0 and
40X − 18Y = 214. Find mean values of X and Y. Solution:
ˉ , Yˉ ), solve the equations simultaneously.
Since regression lines intersect at the mean (X
1. 8X − 10Y = −66
2. 40X − 18Y = 214
Multiply eq(1) by 5: 40X − 50Y = −330
Subtract eq(2) from modified eq(1):

−32Y = −544 ⇒ Y = 17

Substitute Y = 17 in eq(1):

8X − 10(17) = −66 ⇒ 8X − 170 = −66 ⇒ 8X = 104 ⇒ X = 13

ˉ = 13, Yˉ = 17.
Means: X
Example 2: Identifying Equations Given equations: 3X + 2Y − 26 = 0 and 6X + Y − 31 = 0.
Which is X on Y and which is Y on X? Solution:
Assumption: Let 3X + 2Y = 26 be X on Y.
3X = −2Y + 26 ⇒ X = − 23 Y + … ​

bxy = −0.67
​

Then 6X + Y = 31 must be Y on X.
Y = −6X + 31
byx = −6
​

Check: r = bxy × byx =

​ ​
(−0.67)(−6) =
​
4.02 > 1.
​

Since r > 1, Assumption is Wrong.

Correct:
3X + 2Y = 26 is Y on X (2Y = −3X ⋯ ⇒ byx = −1.5). ​

6X + Y = 31 is X on Y (6X = −Y ⋯ ⇒ bxy = −0.16). ​

Check: (−1.5)(−0.16) = ​
0.24 = 0.49. Valid.
​

15. Extra Practice Problems

Theory Questions:
1. Define correlation and distinguish between positive and negative correlation.
2. Explain the properties of Karl Pearson’s coefficient of correlation.
3. What is Rank Correlation? When is it preferred over Pearson’s coefficient?
 4. Distinguish between Correlation and Regression.
5. Why are there two regression lines? Under what condition do they coincide?
6. "Regression coefficients are independent of origin but not of scale." Explain.

Numerical Problems:
1. Find Karl Pearson’s correlation:
X: 10, 12, 18, 24, 23, 27
Y: 13, 18, 12, 25, 30, 10
(Answer: r = 0.25)
2. Calculate Rank Correlation:
X: 15, 20, 28, 12, 40, 60, 20, 80
Y: 40, 30, 50, 30, 20, 10, 30, 60
(Answer: R = -0.32 approx)
3. Find Regression Equations:
X: 1, 2, 3, 4, 5
Y: 2, 5, 3, 8, 7
(Answer: Y on X: Y = 1.1 + 1.3X; X on Y: X = 0.5 + 0.5Y)
ˉ = 40, Yˉ = 60 and bxy = 0.5, byx = 0.8. Find the most likely value of Y when
4. Given means X ​ ​

X=50.
(Answer: Y = 68)
5. Given ∑ X = 30, ∑ Y = 40, ∑ XY = 214, ∑ X 2 = 220, ∑ Y 2 = 340, N = 5. Find r.
(Answer: r = -0.92)

16. Multiple Choice Questions (MCQs)

1. The limits of Karl Pearson’s correlation coefficient are:
A) 0 to 1
B) -1 to 0
C) -1 to +1
D) No limit
Ans: C
2. If r = 0, the variables are:
A) Linearly related
B) Uncorrelated
C) Perfect positive
 D) Perfect negative
Ans: B
3. Rank correlation method was developed by:
A) Karl Pearson
B) R.A. Fisher
C) Spearman
D) Galton
Ans: C
4. Regression coefficient is independent of:
A) Scale
B) Origin
C) Both Scale and Origin
D) Neither
Ans: B
5. If bxy = −0.8 and byx = −0.2, then r is:
​ ​

A) +0.4
B) -0.4
C) +0.16
D) -0.16
Ans: B (Sign of r is same as b)
6. The geometric mean of two regression coefficients is:
A) Mean
B) Correlation coefficient
C) Variance
D) Standard Deviation
Ans: B
7. If correlation is perfect (r = ±1), the angle between regression lines is:
A) 90 degrees
B) 45 degrees
C) 0 degrees
D) 180 degrees
Ans: C
 8. Scatter diagram is used to study:
A) Variability
B) Central Tendency
C) Correlation
D) Skewness
Ans: C
9. If Y decreases as X increases, the correlation is:
A) Positive
B) Negative
C) Zero
D) Linear
Ans: B
10. The regression line of Y on X is derived by minimizing:
A) Vertical distances
B) Horizontal distances
C) Perpendicular distances
D) Both A and B
Ans: A
11. When r = +1, the two regression lines are:
A) Perpendicular
B) Coincident
C) Parallel
D) None of these
Ans: B
12. The value of 1 − r2 is called:
A) Coefficient of determination
B) Coefficient of alienation
C) Standard error
D) Variance
Ans: B (Note: r2 is coeff of determination, 1 − r2 helps find alienation)
13. If one regression coefficient is greater than 1, the other must be:
A) Greater than 1
 B) Less than 1
C) Equal to 1
D) Zero
Ans: B
14. Rank correlation is suitable for:
A) Quantitative data
B) Qualitative data
C) Time series data
D) All of these
Ans: B
15. The arithmetic mean of regression coefficients is:
A) Equal to r
B) Less than r
C) Greater than or equal to r
D) None of these
Ans: C
Notes Created by Noor Mohammed.