𝑚(𝑚2 −1) 2(22 −1) 1

In 𝑥, 68 is repeated twice, then = =

12 12 2

𝑚(𝑚2 −1) 2(22 −1) 1

In 𝑥, 67 is repeated twice, then = =
12 12 2

𝑚(𝑚2 −1) 4(42 −1)

In 𝑦, 67 is repeated 4 times, then = =5
12 12

𝑚(𝑚2 −1) 2(22 −1) 1

In 𝑦, 66 is repeated twice, then = =
12 12 2

𝑚(𝑚2 −1) 2(22 −1) 1

In 𝑦, 65 is repeated twice, then 12
= 12
=2

Rank correlation is

1 1 1 1
6 [∑ 𝑑 2 + 2 + 2 + 5 + 2 + 2]
𝜌 = 1− = 0.722
12(144 − 1)

Linear Regression:
If the variables in bivariate distribution are related, will find that the points in the scatter
diagram will cluster round some curve called the ‘’curve of regression’’. If the curve is a straight
line, it is called the line of regression and there is said to be linear regression between the
variables, otherwise regression is said to be curvilinear.

The lines of regression are the line which gives to be best estimate to the value of one variable
for any specific value of the other variable. Thus, the line of regression is the line of ‘best fit’ and
is obtained by the principle of least squares.

Let us suppose that the in the bivariate distribution (𝑥𝑖 , 𝑦𝑖 ); 𝑖 = 1,2,3, … , 𝑛; 𝑦 is dependent
variable and 𝑥 is independent variable. Let the line of regression is the line of 𝑦 on 𝑥 be

𝑦 = 𝑎 + 𝑏𝑥 (1)

Eq. (1) represents the family of straight lines for different values of the arbitrary constants
′𝑎′ 𝑎𝑛𝑑 ′𝑏′. The problem is to determine the ′𝑎′ 𝑎𝑛𝑑 ′𝑏 so that the line Eq. (1) is the line of best
fit.
According to the principle of the principle of least squares, we have to determine ′𝑎′ 𝑎𝑛𝑑 ′𝑏′.
𝑛
2
𝐸 = ∑(𝑦𝑖 − (𝑎 + 𝑏𝑥𝑖 ))
𝑖=1

Is minimum. From the principle of maxima and minima, the partial derivatives of 𝐸, with respect
to ′𝑎′ 𝑎𝑛𝑑 ′𝑏′ should vanish separately, i.e.,
𝑛
𝜕𝐸
= 0 = −2 ∑(𝑦𝑖 − (𝑎 + 𝑏𝑥𝑖 ))
𝜕𝑎
𝑖=1

∑𝑛𝑖=1 𝑦𝑖 = 𝑎𝑛 + ∑𝑛𝑖=1 𝑦𝑖 = 𝑎𝑛 + 𝑏 ∑𝑛𝑖=1 𝑥𝑖 (2)

𝑛
𝜕𝐸
= 0 = −2 ∑ 𝑥𝑖 (𝑥𝑖 − (𝑎 + 𝑏𝑥𝑖 ))
𝜕𝑏
𝑖=1

∑𝑛𝑖=1 𝑥𝑖 𝑦𝑖 = 𝑎 ∑𝑛𝑖=1 𝑥𝑖 + 𝑏 ∑𝑛𝑖=1 𝑥𝑖 2 (3)

Dividing on both sides to the Eq. (2) by 𝑛, we get

𝑦̅ = 𝑎 + 𝑏𝑥̅ (4)

Now, the line of regression of 𝑌 𝑜𝑛 𝑋 passes through the point (𝑥̅ , 𝑦̅ ).

1
𝜇11 = 𝐶𝑜𝑣(𝑥, 𝑦) = 𝑛 ∑𝑛𝑖=1 𝑥𝑖 𝑦𝑖 − 𝑥̅ 𝑦̅

1
∑𝑛𝑖=1 𝑥𝑖 𝑦𝑖 = 𝜇11 + 𝑥̅ 𝑦̅ (5)
𝑛

𝑛
2
1
𝜎𝑥 = ∑ 𝑥𝑖 2 − 𝑥̅ 2
𝑛
𝑖=1

1
∑𝑛𝑖=1 𝑥𝑖 2 = 𝜎𝑥 2 + 𝑥̅ 2 (6)
𝑛

Dividing Eq. (3) by 𝑛 and using Eqs. (5) and (6), we get

𝜇11 + 𝑥̅ 𝑦̅ = 𝑎𝑥̅ + 𝑏(𝜎𝑥 2 + 𝑥̅ 2 ) (7)

Eq. (7)- Eq. (4)× 𝑥̅ , we get

 𝜇11 = 𝑏𝜎𝑥 2

𝜇11
𝑏=
𝜎𝑥 2

Since ‘b’ is the slope of the line of regression of Y on X and since the line of regression passes
through the point (𝑥̅ , 𝑦̅ ) its equation is

𝜇11
𝑌 − 𝑦̅ = 𝑏(𝑥 − 𝑥̅ ) = (𝑋 − 𝑥̅ )
𝜎𝑥 2

𝜎𝑌
𝑌 − 𝑦̅ = 𝑟 (𝑋 − 𝑥̅ )
𝜎𝑋

Starting the equation 𝑋 = 𝐴 + 𝐵𝑌 and proceeding similarly, we get

𝜎𝑋
𝑋 − 𝑥̅ = 𝑟 (𝑌 − 𝑦̅)
𝜎𝑌

Problem:

From the following data, obtain the two regression equations

Sales 91 97 108 121 67 124 51 73 111 57

Purchases 71 75 69 97 70 91 39 61 80 47

Solution:

Let us denote the sales by the variable 𝑥 𝑎𝑛𝑑 𝑦 the purchases by the variable 𝑦
𝑥 𝑦 𝑑𝑥 𝑑𝑦 𝑑𝑥 2 𝑑𝑦 2 𝑑𝑥 𝑑𝑦
= 𝑥 − 90 = 𝑦 − 70
91 71 1 1 1 1 1
97 75 7 5 49 25 35
108 69 18 -1 324 1 -18
121 97 31 27 961 729 837
67 70 -23 0 529 0 0
124 91 34 21 1156 441 714
51 39 -39 -31 1521 961 1209
73 61 -17 -9 289 81 153
111 80 21 10 441 100 210
 57 47 -33 -23 1089 529 759
∑𝑥 ∑𝑦 ∑ 𝑑𝑥 = 0 ∑ 𝑑𝑦 = 0 ∑ 𝑑𝑥 2 ∑ 𝑑𝑦 2 ∑ 𝑑𝑥 𝑑𝑦
= 900 = 700 = 6360 = 2868 = 3900

∑𝑥 900
We have, 𝑥̅ = = = 90
𝑛 10

∑ 𝑦 700
𝑦̅ = = = 70
𝑛 10
∑(𝑥 − 𝑥̅ )(𝑦 − 𝑦̅) ∑ 𝑑𝑥𝑑𝑦 3900
𝑏𝑦𝑥 = = = = 0.6132
∑(𝑥 − 𝑥̅ )2 ∑ 𝑑𝑥 2 6360
∑(𝑥 − 𝑥̅ )(𝑦 − 𝑦̅) ∑ 𝑑𝑥𝑑𝑦 3900
𝑏𝑥𝑦 = = = = 1.361
∑(𝑦 − 𝑦̅)2 ∑ 𝑑𝑦 2 2868
Equation of regression of 𝑦 𝑜𝑛 𝑥 is
𝑦 − 𝑦̅ = 𝑏𝑦𝑥 (𝑥 − 𝑥̅ )

𝑦 − 70 = 0.6132(𝑥 − 90)
𝑦 − 70 = 0.6132 𝑥 − 0.613 × 90
= 0.6132 𝑥 − 55.188
𝑦 = 0.6132 𝑥 − 55.188 + 70
𝑦 = 0.6132 𝑥 + 14.812

Equation of regression of 𝑥 𝑜𝑛 𝑦 is
𝑥 − 𝑥̅ = 𝑏𝑥𝑦 (𝑦 − 𝑦̅)

𝑥 − 90 = 1.361(𝑦 − 70)
𝑥 − 90 = 1.361 𝑦 − 1.361 × 70
= 1.361 𝑦 − 95.27
𝑥 = 1.361 𝑦 − 95.27 + 90
𝑥 = 1.361 𝑦 − 5.27
𝑟 2 = 𝑏𝑦𝑥 . 𝑏𝑥𝑦

𝑟 2 = 0.6132 × 1.361 = 0.8346

𝑟 = ∓0.9135
But since, both the regression coefficients are positive, 𝑟 must be positive.
𝑟 = 0.9135
Problem:

From the data given below find

(a) Two regression coefficients
(b) The two regression equations
(c) The coefficient of correlation between the marks in Economics and Statistics
(d) The most likely marks in Statistics when marks in Economics are 30.

Marks in 25 28 35 32 31 36 29 38 34 32
Economics
Marks in 43 46 49 41 36 32 31 30 33 39
Statistics

Solution:
𝑥 𝑦 𝑑𝑥 𝑑𝑦 𝑑𝑥 2 𝑑𝑦 2 𝑑𝑥 𝑑𝑦
= 𝑥 − 32 = 𝑦 − 38
25 43 -7 5 49 25 -35
28 46 -4 8 16 64 -32
35 49 3 11 9 121 33
32 41 0 3 0 9 0
31 36 -1 -2 1 4 2
36 32 4 -6 16 36 -24
29 31 -3 -7 9 49 21
38 30 6 -8 36 64 -48
34 33 2 -5 4 25 -10
32 39 0 1 0 1 0
∑𝑥 ∑𝑦 ∑ 𝑑𝑥 = 0 ∑ 𝑑𝑦 = 0 ∑ 𝑑𝑥 2 ∑ 𝑑𝑦 2 ∑ 𝑑𝑥 𝑑𝑦
= 320 = 380 = 140 = 398 = −93

∑ 𝑥 320
𝑥̅ = = = 32
𝑛 10
∑ 𝑦 380
𝑦̅ = = = 38
𝑛 10
(a) Regression coefficients:

Coefficient of regression 𝑦 𝑜𝑛 𝑥 is given by

 ∑(𝑥 − 𝑥̅ )(𝑦 − 𝑦̅) ∑ 𝑑𝑥𝑑𝑦 −93
𝑏𝑦𝑥 = = = = −0.6643
∑(𝑥 − 𝑥̅ )2 ∑ 𝑑𝑥 2 140
∑(𝑥 − 𝑥̅ )(𝑦 − 𝑦̅) ∑ 𝑑𝑥𝑑𝑦 −93
𝑏𝑥𝑦 = = = = −0.2337
∑(𝑦 − 𝑦̅)2 ∑ 𝑑𝑦 2 398
(b) Equations of the line of regression of 𝑥 𝑜𝑛 𝑦 is
𝑥 − 𝑥̅ = 𝑏𝑥𝑦 (𝑦 − 𝑦̅)

𝑥 − 32 = −0.2337(𝑦 − 38)
𝑥 − 32 = −0.2337 𝑦 + 38 × 0.2337
= −0.2337 𝑦 + 8.8806
𝑥 = −0.2337 𝑦 + 8.8806 + 32
𝑥 = −0.2337 𝑦 + 40.8806
Equation of line of regression of 𝑦 𝑜𝑛 𝑥 is
𝑦 − 𝑦̅ = 𝑏𝑦𝑥 (𝑥 − 𝑥̅ )

𝑦 − 38 = −0.6643(𝑥 − 32)
𝑦 − 38 = −0.6643 𝑥 + 0.6643 × 32
= −0.6643 𝑥 + 0.6643 × 32 + 38
𝑦 = −0.6643 𝑥 + 59.2576
(c) Correlation coefficient:
𝑟 2 = 𝑏𝑦𝑥 . 𝑏𝑥𝑦

𝑟 2 = (−0.6643)(−0.2337) = 0.1552
𝑟 = ∓0.394
Since the both regression coefficients are negative. Hence the discarding plus sign, we get
𝑟 = −0.394

(d) In order to estimate the most likely marks in Statistics (𝑦) when marks in Economics (𝑥)
are 30, we use the line of regression of 𝑦 𝑜𝑛 𝑥.
The equation is
𝑦 − 38 = −0.6643 (30) + 59.2576
𝑦 = 39.3286
Hence the most likely marks in Statistics when in Economics are 30, are 39.3286≈ 39.

Problem:

The following is an estimated supply regression for sugar:

𝑦 = 0.025 + 1.5 𝑥, where 𝑦 is supply in kilos and 𝑥 is price in rupees per kilo.
(a) Interpret the coefficient of variable 𝑥
(b) Predict the supply when supply when price is Rs. 20 per kilo
(c) Given that 𝑟(𝑥, 𝑦) = 1, interpret the implied relationship between price and quality
supplied.
Solution:
The regression equation of 𝑦 (supply in kgs) on 𝑥 (price in rupees per kg) is given to be
𝑦 = 0.025 + 1.5 𝑥 = 𝑎 + 𝑏𝑥 (say) (1)
(a) The coefficients of variation 𝑥
𝑏 = 1.5 is the coefficient of regression of 𝑦 𝑜𝑛 𝑥. It reflects the unit change in the value
of 𝑦, for a unit change in the corresponding value of 𝑥. This means that if the price of
sugar goes up by Re. 1 per kg, the estimated supply of sugar goes up by 1.5 kg.

(b) From eq. (1), the estimated supply of sugar when its price is Rs. 20 per kg is given by
𝑦 = 0.025 + 1.5 × 20 = 30.025 kg
(c) 𝑟(𝑥, 𝑦) = 1
The relationship between that 𝑥 𝑎𝑛𝑑 𝑦 is exactly linear. i.e., all the observed values (𝑥, 𝑦)
lies on straight line.
Problem:

Given that the regression equations of 𝑦 𝑜𝑛 𝑥 and of 𝑥 𝑜𝑛 𝑦 are respectively 𝑦 = 𝑥 and

4𝑥 − 𝑦 = 3, and that the second moment of 𝑥 about the origin is 2, find
(a) The correlation coefficient between 𝑥 𝑎𝑛𝑑 𝑦
(b) The standard deviation of 𝑦

Solution:
Regression equation of 𝑦 𝑜𝑛 𝑥 is 𝑦 = 𝑥
 𝑏𝑦𝑥 = 1

Regression equation of 𝑥 𝑜𝑛 𝑦 is 4𝑥 − 𝑦 = 3
1 3
𝑥= 𝑦+
4 4
1
𝑏𝑥𝑦 =
4

(a) The correlation coefficient between 𝑥 𝑎𝑛𝑑 𝑦 is

𝑟 2 = 𝑏𝑦𝑥 . 𝑏𝑥𝑦
1 1
𝑟2 = 1 × =
4 4
𝑟 = ∓0.5
Since the both the regression coefficients are positive 𝑟 = 0.5.
∑ 𝑥2
(b) We are given that the second moment of 𝑥 about origin is 2. i.e., 𝑛
=2

Since (𝑥̅ , 𝑦̅) is the point of intersection of the two lines of regression
Solving 𝑦 = 𝑥 and 4𝑥 − 𝑦 = 3, then 𝑥 = 1 = 𝑦
𝑥̅ = 1 𝑎𝑛𝑑 𝑦̅ = 1

∑ 𝑥2
𝜎𝑥 2 = − 𝑥̅ 2 = 2 − 1 = 1
𝑛
𝜎𝑦
Also, 𝑏𝑦𝑥 = 𝑟
𝜎𝑥

1 𝜎𝑦
1= ( )
2 1
𝜎𝑦 = 2
Coefficient of Determination:
Coefficient of correlation between two variable series is a measure of linear relationship between
them and indicates the amount of variation of one variable which is associated with or accounted
for by another variable. A more useful and readily comprehensible measure for this purpose is
the coefficient of determination which gives the percentage variation in the dependent variable
that is accounted for by the independent variable.
In other words, the coefficient of determination gives the ratio of the explained variance to the
total variance. The coefficient is given by the square of the correlation coefficient i.e.,
𝑒𝑥𝑝𝑙𝑎𝑖𝑛𝑒𝑑 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒
𝑟2 = .
𝑡𝑜𝑡𝑎𝑙 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒

Ex:
If the value of 𝑟 = 0.8, we cannot conclude that 80% of the variation in the relative series
(dependent variable) is due to the variation in the subject series (independent variable). But the
coefficient of determination in this case 𝑟 2 = 0.64 which implies that only 64% of the variation
in the relative series has been explained by the subject series and the remaining 36% of the
variation is due to other factors.

Coefficient of Partial correlation:

Sometimes the correlation between two variables 𝑿𝟏 𝒂𝒏𝒅 𝑿𝟐 may be partly due to the
correlation of third variable 𝑿𝟑 with both 𝑿𝟏 𝒂𝒏𝒅 𝑿𝟐 . In such a situation, one may want to
know what the correlation between 𝑋1 𝑎𝑛𝑑 𝑋2 would be if the effect of 𝑋3 an each of 𝑋1 𝑎𝑛𝑑 𝑋2
were eliminated. This correlation is called partial correlation and the correlation coefficient
between 𝑿𝟏 𝒂𝒏𝒅 𝑿𝟐 after the linear effect of 𝑿𝟑 on each of them has been eliminated is
called the partial coefficient.
The partial correlation coefficient between 𝑋1 𝑎𝑛𝑑 𝑋2, usually denoted by 𝑟12.3 is given by
𝑟12 − 𝑟13 𝑟23
𝑟12.3 =
√(1 − 𝑟13 2 )(1 − 𝑟23 2 )

Similarly,
 𝑟13 − 𝑟12 𝑟32
𝑟13.2 =
√(1 − 𝑟12 2 )(1 − 𝑟32 2 )
and
𝑟23 − 𝑟21 𝑟31
𝑟23.1 =
√(1 − 𝑟21 2 )(1 − 𝑟31 2 )

Multiple correlation in terms of total and partial correlations:

2 𝑟12 2 + 𝑟13 2 − 2𝑟12 𝑟13 𝑟23

1 − 𝑅1.23 =1−
1 − 𝑟23 2
1 − 𝑟23 2 − 𝑟12 2 − 𝑟13 2 + 2𝑟12 𝑟13 𝑟23
=
1 − 𝑟23 2

Note:
𝜔
1 − 𝑅1.23 2 =
𝜔11

1 𝑟12 𝑟13
Where, 𝜔 = |𝑟21 1 𝑟23 | = 1 − 𝑟12 2 − 𝑟13 2 − 𝑟23 2 + 2𝑟12 𝑟13 𝑟23
𝑟31 𝑟32 1
1 𝑟23
and 𝜔11 = | | = 1 − 𝑟23 2
𝑟32 1

Problem:
From the data relating to the yield of dry bark (𝑋1 ), height (𝑋2 ) and girth (𝑋3 ) for 18 cinchona
plants, the following correlation coefficients were obtained:
𝑟12 = 0.77, 𝑟13 = 0.72 𝑎𝑛𝑑 𝑟23 = 0.52. find the partial correlation coefficients 𝑟12.3 and multiple
correlation coefficient 𝑅1.23.
Solution:

𝑟12 − 𝑟13 𝑟23

𝑟12.3 =
√(1 − 𝑟13 2 )(1 − 𝑟23 2 )
 0.77 − 0.72 × 0.52
= = 0.62
√(1 − 0.722 )(1 − 0.522 )

𝑟12 2 + 𝑟13 2 − 2𝑟12 𝑟13 𝑟23

𝑅1.23 2 =
1 − 𝑟23 2
0.772 + 0.722 − 2 × 0.77 × 0.72 × 0.52
= = 0.7334
1 − 0.522
𝑅1.23 = +0.8564 (since multiple correlation is non-negative).

Problem:
In a trivariate distribution 𝜎1 = 2, 𝜎2 = 𝜎3 = 3, 𝑟12 = 0.7, 𝑟23 = 𝑟31 = 0.5.
Find (i) 𝑟23.1 (ii) 𝑅1.23 (iii) 𝑏12.3 , 𝑏13.2 and (iv) 𝜎1.23 .
Solution:

𝑟23 − 𝑟21 𝑟31

𝑟23.1 =
√(1 − 𝑟21 2 )(1 − 𝑟31 2 )
0.5 − 0.7 × 0.5
= = 0.2425
√(1 − 0.72 )(1 − 0.52 )
(ii)
𝑟12 2 + 𝑟13 2 − 2𝑟12 𝑟13 𝑟23
𝑅1.23 2 =
1 − 𝑟23 2
0.72 + 0.52 − 2 × 0.7 × 0.5 × 0.5
𝑅1.23 2 = = 0.52
1 − 0.52

𝑅1.23 = +0.7211
(iii)
𝜎 𝜎
𝑏12.3 = 𝑟12.3 (𝜎1.3 ) and 𝑏13.2 = 𝑟13.2 (𝜎1.2 ) (1)
2.3 3.2

𝑟12 − 𝑟13 𝑟23

𝑟12.3 = = 0.6
√(1 − 𝑟13 2 )(1 − 𝑟23 2 )
 𝑟13 − 𝑟12 𝑟32
𝑟13.2 = = 0.2425
√(1 − 𝑟12 2 )(1 − 𝑟32 2 )

𝜎1.3= 𝜎1 √(1 − 𝑟13 2 ) = 2 × √(1 − 0.52 ) = 1.7320

𝜎2.3= 𝜎2 √(1 − 𝑟23 2 ) = 3 × √(1 − 0.52 ) = 2.5980

𝜎1.2= 𝜎1 √(1 − 𝑟12 2 ) = 2 × √(1 − 0.72 ) = 1.4282

𝜎3.2= 𝜎3 √(1 − 𝑟32 2 ) = 2 × √(1 − 0.52 ) = 2.5980

1.7320 1.4282
Eq. (1) gives 𝑏12.3 = 0.6 × = 0.4 and 𝑏13.2 = 0.2425 × = 0.1333
2.5980 2.5980

𝜔
(iv) 𝜎1.23 = 𝜎1 (√𝜔 )
11

1 𝑟12 𝑟13
𝑟
𝜔 = | 21 1 𝑟23 | = 1 − 𝑟12 2 − 𝑟13 2 − 𝑟23 2 + 2𝑟12 𝑟13 𝑟23 = 0.36
𝑟31 𝑟32 1
1 𝑟23
and 𝜔11 = | | = 1 − 𝑟23 2 = 1 − 0.52 = 0.75
𝑟32 1

0.36
𝜎1.23 = 2 × (√ ) = 1.3856
0.75
Multiple regression:

Bivariate Regression equation:

 Here we try to study the linear relationship between two variables x and y.
Y = a + bX = β0 + 𝛽1 𝑋
We see that the 𝑎 = 𝛽0 is the Y intercept, 𝑏 = 𝛽1 is the slope of the linear relationship between
the variable X and Y.
Multivariate regression equation
 Y = a + b1X1 + b2X2 = β0 + β1X1 + β2X2
 b1 = β1 = partial slope of the linear relationship between the first independent variable and
Y, indicates the change in Y for one unit change in X1.
 b2 = β2 = partial slope of the linear relationship between the second independent variable
and Y, indicates the change in Y for one unit change in X2.

Formulas for finding partial slopes:

𝑆𝑦 𝑟𝑦1 − 𝑟𝑦2 𝑟12
𝑏1 = 𝛽1 = ( 2 )
𝑆1 1 − 𝑟12
𝑆𝑦 𝑟𝑦2 − 𝑟𝑦1 𝑟12
𝑏2 = 𝛽2 = ( 2 )
𝑆2 1 − 𝑟12

Sy= standard deviation of Y

S1 = standard deviation of the first independent variable (X1)
S2 = standard deviation of the second independent variable (X2)
ry1 = bivariate correlation between Y and X1
ry2 = bivariate correlation between Y and X2
r12 = bivariate correlation between X1 and X2

Example:
1) The salary of a person in an organisation has to be regressed in terms of experience (X1)
and mistakes (X2). If it is given that the values
𝑌̅ = 3.3; 𝑋
̅̅̅1 = 2.7; 𝑋
̅̅̅2 = 13.7
𝑆𝑦 = 2.1; 𝑆1 = 1.5; 𝑆2 = 2.6
and the zero order correlations :
𝑟𝑦1 = 0.5; 𝑟𝑦2 = −0.3; 𝑟12 = −0.47;
Find the linear regression and interpret the results.

So,
𝑆𝑦 𝑟𝑦1 − 𝑟𝑦2 𝑟12
𝑏1 = 𝛽1 = ( 2 )
𝑆1 1 − 𝑟12
2.1 0.50 − (−0.3)(−0.47)
𝑏1 = 𝛽1 = ( ) = 0.65
1.5 1 − (−0.47)2
Similarly,
𝑆𝑦 𝑟𝑦2 − 𝑟𝑦1 𝑟12
𝑏2 = 𝛽2 = ( 2 )
𝑆2 1 − 𝑟12
2.1 0.30 − (0.5)(−0.47)
𝑏2 = 𝛽2 = ( ) = −0.07
2.6 1 − (−0.47)2
Calculation of a:
𝑎 = 𝑌̅ − 𝑏1 𝑋
̅̅̅1 − 𝑏2 ̅̅̅
𝑋2
𝛽0 = 𝑌̅ − 𝛽1 𝑋̅̅̅1 − 𝛽2 𝑋
̅̅̅2
= 3.3 − (0.65)(2.7) − (−0.07) 13.7
= 2.5
Interpretation:
1) If a person has no experience and has not done any mistakes, he would get a salary of
2.5 units.
2) If the experience goes up by 1 unit, there would be an increment in the salary by 0.65
units.
3) If he/ she commits a mistake, then the salary would decrease by 0.07 units.
 Module-4

DISCRETE PROBABILITY DISTRIBUTIONS

Bernoulli’s trials:
Suppose, associated with random trial there is an event called ‘success’ and the complementary
event is called ‘failure’. Let the probability for success be 𝑝 and probability for failure be 𝑞.
Suppose the random trials are prepared 𝑛 times under identical conditions. These are called
Bernoullian trials.

Bernoulli’s Distribution:

A random variable 𝑋 which takes two values 0 and 1 with probability 𝑞 𝑎𝑛𝑑 𝑝 respectively. That
is 𝑃(𝑋 = 0) = 𝑞 𝑎𝑛𝑑 𝑃(𝑋 = 1) = 𝑝, 𝑞 = 1 − 𝑝 is called a Bernoulli’s discrete random
variable. The probability function of Bernoulli’s distribution can be written as

𝑃(𝑋) = 𝑝 𝑋 𝑞 𝑛−𝑋 = 𝑝 𝑋 (1 − 𝑝)𝑛−𝑋 ; 𝑋 = 0,1

Note:

1. Mean of Bernoulli’s distribution discrete random variable 𝑋

𝜇 = 𝐸(𝑋) = ∑ 𝑋𝑖 . 𝑃(𝑋𝑖 ) = (0 × 𝑞) + (1 × 𝑝) = 𝑝

2. Variance of 𝑋 is

𝑉(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋)2 = ∑ 𝑋𝑖 2 𝑃(𝑋𝑖 ) − 𝜇 2

= (02 × 𝑞) + (12 × 𝑝) − 𝑝2 = 𝑝 − 𝑝2 = 𝑝(1 − 𝑝) = 𝑝𝑞

The standard deviation is 𝜎 = √𝑝𝑞

Probability Binomial Distribution:

Let a random experiment be performed repeatedly and let the occurrence of an event 𝐴 is any
trial be called success and non-occurrence 𝑃(𝐴̅), a failure (Bernoulli trial). Consider a series of 𝑛
independent Bernoulli trials (𝑛 being finite) in which the probability of success 𝑃(𝐴) = 𝑝 or
𝑃(𝐴̅) = 1 − 𝑝 = 𝑞 in any trial is constant for each trial.

𝑃(𝑋 = 𝑥) = 𝑛𝐶𝑥 𝑝 𝑥 𝑞 𝑛−𝑥 , 𝑥 = 0,1,2,3, … , 𝑛

Since the probabilities of 0,1,2, 3,…,n successes, namely 𝑞 𝑛 , 𝑛𝐶1 𝑞 𝑛−1 𝑝, 𝑛𝐶2 𝑞 𝑛−2 𝑝2 , … , 𝑝𝑛 are
the successive terms of the Binomial expansion of (𝑞 + 𝑝)𝑛 , the probability distribution so
obtained is called Binomial probability distribution.

Definition:

A random variable 𝑋 is said to be follow Binomial distribution denoted by B (n, p), if it assumes
only non-negative values and its probability mass function is given by
𝑛𝐶𝑥 𝑝 𝑥 𝑞 𝑛−𝑥 , 𝑥 = 0,1,2,3, … , 𝑛
𝑃(𝑋 = 𝑥) = {
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Where 𝑛 𝑎𝑛𝑑 𝑝 are known as parameters.
Note:
 If 𝑛 is also sometimes known as the degree of the distribution
 ∑𝑛𝑥=0 𝑛𝐶𝑥 𝑝 𝑥 𝑞 𝑛−𝑥 = (𝑞 + 𝑝)𝑛 = 1
 The Binomial distribution is important not only because of its wide range applicability,
but also because it gives rise to many other probability distributions.
 Any variable which follows Binomial distribution is known as Binomial variate.
Conditions for Binomial Experiment:

The Bernoulli process involving a series of independent trials, is based on certain conditions as
under:

 There are only two mutually exclusive and collective exhaustive outcomes of the random
variable and one of them is referred to as a success and the other as a failure.
 The random experiment is performed under the same conditions for a fixed and finite
(also discrete) number of times, say n. Each observation of the random variable in a
random experiment is called a trial. Each trial generates either a success denoted by 𝑝 or
a failure denoted by 𝑞.
 The outcome (i.e., success or failure) of any trial is not affected by the outcome of any
other trial.
 All the observations are assumed to be independent of each of each other. This means
that the probability of outcomes remains constant throughout the process.

Example:

To understand the Bernoulli process, consider the coin tossing problem where 3 coins are tossed.
Suppose we are interested to know the probability of two heads. The possible sequence of
outcomes involving two heads can be obtained in the following three ways:
HHT, HTH, THH.

Binomial Probability Function:

In general, for a binomial random variable, 𝑋 the probability of success (occurrence of desired
outcome) 𝑟 number of times in 𝑛 independent trials, regardless of their order of occurrence is
given by the formula:
𝑛!
𝑃(𝑋 = 𝑟 𝑠𝑢𝑐𝑐𝑒𝑠𝑠𝑒𝑠) = 𝑛𝐶𝑟 𝑝𝑟 𝑞 𝑛−𝑟 = 𝑝𝑟 𝑞 𝑛−𝑟 , 𝑟 = 0,1,2,3, … , 𝑛
(𝑛 − 𝑟)! 𝑟!

where
n = number of trials (specified in advance) or sample size
p = probability of success
q = (1 – p), probability of failure
x = discrete binomial random variable
r = number of successes in n trials

Relationship between mean and variance:

Mean of a Binomial distribution:
The Binomial probability distribution is given by
𝑝(𝑟) = 𝑛𝐶𝑟 𝑝𝑟 𝑞 𝑛−𝑟 ; 𝑟 = 0,1,2, … , 𝑛, 𝑎𝑛𝑑 𝑞 − 1 − 𝑝

Mean of 𝑋 is
𝑛 𝑛

𝜇 = 𝐸(𝑋) = ∑ 𝑟 𝑝(𝑟) = ∑ 𝑟 𝑛𝐶𝑟 𝑝𝑟 𝑞 𝑛−𝑟

𝑟=0 𝑟=0

= 0 × 𝑞 𝑛 + 1 × 𝑛𝐶1 𝑝𝑞 𝑛−1 + 2 × 𝑛𝐶2 𝑝2 𝑞 𝑛−2 + 3 × 𝑛𝐶3 𝑝3 𝑞 𝑛−3 + ⋯ + 𝑛 𝑝𝑛

𝑛(𝑛 − 1) 2 𝑛−2 𝑛(𝑛 − 1)(𝑛 − 2) 3 𝑛−3
= 𝑛𝑝𝑞 𝑛−1 + 2. 𝑝 𝑞 + 3. 𝑝 𝑞 + ⋯ + 𝑛 𝑝𝑛
2! 3!
(𝑛 − 1)(𝑛 − 2) 2 𝑛−3
= 𝑛𝑝 [𝑞 𝑛−1 + (𝑛 − 1)𝑝𝑞 𝑛−2 + 𝑝 𝑞 + ⋯ + 𝑝𝑛−1 ]
2!
= 𝑛𝑝(𝑞 + 𝑝)𝑛−1
= 𝑛𝑝
𝜇 = 𝐸(𝑋) = 𝑛𝑝

Variance of a Binomial distribution:

Variance 𝑉(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋)2
𝑛

= ∑ 𝑟 2 𝑝(𝑟) − 𝜇 2
𝑟=0
𝑛

= ∑[𝑟(𝑟 − 1) + 𝑟]. 𝑝(𝑟) − 𝜇 2

𝑟=0
 𝑛 𝑛
𝑟 𝑛−𝑟
= ∑ 𝑟(𝑟 − 1) 𝑛𝐶𝑟 𝑝 𝑞 + ∑ 𝑟 𝑝(𝑟) − 𝜇 2
𝑟=0 𝑟=0

= [2. 𝑛𝐶2 𝑝2 𝑞 𝑛−2 + 3.2. 𝑛𝐶3 𝑝3 𝑞 𝑛−3 + ⋯ + 𝑛(𝑛 − 1) 𝑝𝑛 ] + 𝜇 − 𝜇 2

𝑛(𝑛 − 1) 2 𝑛−2 𝑛(𝑛 − 1)(𝑛 − 2) 3 𝑛−3
= [2. 𝑝 𝑞 + 6. 𝑝 𝑞 + ⋯ + 𝑛(𝑛 − 1) 𝑝𝑛 ] + 𝜇 − 𝜇 2
2! 3!
(𝑛 − 2)(𝑛 − 3) 2 𝑛−4
= 𝑛(𝑛 − 1)𝑝2 [𝑞 𝑛−2 + (𝑛 − 2)𝑝𝑞 𝑛−3 + 𝑝 𝑞 + ⋯ + 𝑝𝑛−2 ] + 𝜇 − 𝜇 2
2!

= 𝑛(𝑛 − 1)𝑝2 (𝑞 + 𝑝)𝑛−2 + 𝜇 − 𝜇 2

= 𝑛(𝑛 − 1)𝑝2 + 𝑛𝑝 − (𝑛𝑝)2
= 𝑛2 𝑝2 − 𝑛𝑝2 + 𝑛𝑝 − 𝑛2 𝑝2 = 𝑛𝑝 − 𝑛𝑝2 = 𝑛𝑝(1 − 𝑝) = 𝑛𝑝𝑞
𝑉(𝑋) = 𝑛𝑝𝑞

Problem 1:
A fair coin is tossed six times, then find the probability of getting four heads.
Solution:
𝑝= probability of getting a head=1/2
𝑞= probability of not getting a head=1/2
𝑛 = 6, 𝑟 = 4
𝑝(𝑟) = 6𝐶4 𝑝𝑟 𝑞 𝑛−𝑟

1 4 1 2
𝑝(4) = 6𝐶4 ( ) ( )
2 2
6! 1 6 15
= ( ) =
4! 2! 2 64

Problem 2:

The incidence of an occupational disease in an industry is such that the workers have a 20%
chance of suffering from it. What is the probability that out of 6 workers chosen at random, four
or more will suffer from disease?
Solution:

The probability of a worker suffering from disease= p=20%=0.2

The probability that of no worker suffering from disease= q=80%=0.8

The probability that four or more workers suffer from disease = 𝑃(𝑋 ≥ 4)
𝑃(𝑋 ≥ 4) = 𝑃(𝑋 = 4) + 𝑃(𝑋 = 5) + 𝑃(𝑋 = 6)

= 6𝐶4 (0.2)4 (0.8)2 + 6𝐶5 (0.2)5 (0.8) + 6𝐶6 (0.2)6 = 0.0175

Problem 3:

Six dice are thrown 729 times. How many times do you except at least three dice to show a 5 or
6.

Solution:
2 1
𝑝 = probability of occurrence of 5 or 6 in one throw= 6 = 3
1 2
𝑞 =1−𝑝=1− =
3 3
𝑛=6
The probability of getting at least three dice to show a 5 or 6
𝑃(𝑋 ≥ 3) = 𝑃(𝑋 = 3) + 𝑃(𝑋 = 4) + 𝑃(𝑋 = 5) + 𝑃(𝑋 = 6)
1 3 2 3 1 4 2 2 1 5 2 1 1 6
= 6𝐶3 ( ) ( ) + 6𝐶4 ( ) ( ) + 6𝐶5 ( ) ( ) + 6𝐶6 ( )
3 3 3 3 3 3 3
1 233
= (160 + 60 + 12 + 1) =
(3) 6 729
The expected number of such cases in 729 times
233
= 729 ( ) = 233
729
Problem 4:

If the probability of a detective bolt is 0.2, find

(i) Mean and
(ii) Standard deviation for the bolts in a total of 400.

Solution:
Given 𝑛 = 400, 𝑝 = 0.2, 𝑞 = 0.8
(i) Mean is 𝑛𝑝 = 400 × 0.2 = 80
(ii) Standard deviation is √𝑛𝑝𝑞 = √80 × 0.8 = √64 = 8

Problem 5:
Find the maximum 𝑛 such that the probability of getting no head in tossing a fair coin 𝑛 times is
greater than 0.1.

Solution:

𝑝 = probability of getting a head = ½

𝑞 = probability of not getting a head = 1- ½ = ½
Probability of getting no head in tossing a fair coin 𝑛 times is greater than 0.1 is
𝑃(𝑋 = 0) > 0.1
𝑛𝐶0 𝑝0 𝑞 𝑛 > 0.1
𝑞 𝑛 > 0.1
1 𝑛
( ) > 0.1
2
2𝑛 < 10, 𝑡ℎ𝑒𝑛 𝑛 > 3.
Hence the required maximum 𝑛 = 3.
Problem 6:
Fit a binomial distribution to the following frequency distribution
𝑥 0 1 2 3 4 5 6
𝑓 13 25 52 58 32 16 4

Solution:
The number of trials is 𝑛 = 6
𝑁 = ∑ 𝑓𝑖 = total frequency
∑ 𝑓 𝑖 𝑥𝑖 25+104+174+128+8+24
Mean = ∑ 𝑓𝑖
= 200
= 2.675

Mean = 𝑛𝑝
2.675
𝑛𝑝 = 6𝑝, 𝑡ℎ𝑒𝑛 𝑝 = = 0.446
6
𝑞 = 1 − 0.446 = 0.554

Binomial distribution to be fitted is 𝑁(𝑞 + 𝑝)𝑛 = 200(0.554 + 0.446)6

= 200[6𝐶0 (0.554)6 + 6𝐶1 (0.554)5 (0.446) + 6𝐶2 (0.554)4 (0.446)2 + 6𝐶3 (0.554)3 (0.446)3
+ 6𝐶4 (0.554)2 (0.446)4 + 6𝐶5 (0.554)1 (0.446)5 + 6𝐶6 (0.446)6 ]
= 200[0.02891 + 0.1396 + 0.2809 + 0.3016 + 0.1821 + 0.05864 + 0.007866]
= 5.782 + 27.92 + 56.18 + 60.32 + 36.42 + 11.728 + 1.5732
The successive terms in the expansion give the expected or theoretical frequencies which are

𝑥 0 1 2 3 4 5 6

𝑓 6 28 56 60 36 12 2
(expected
or
theoretical
frequencies)
 Home Work:

1. A die is tossed thrice. A success is getting 1 or 6 on a toss. Find the mean and variance of
the number of successes.
2. The mean and variance of a binomial distribution are 4 and 4/3 respectively. Then find
𝑃(𝑋 ≥ 1).
3. Fit a binomial distribution to the following frequency distribution
𝑥 0 1 2 3 4 5
𝑓 2 14 20 34 22 8

Moment Generating Function of Binomial distribution:

Let 𝑋~𝐵(𝑛, 𝑝), then

𝑀(𝑡) = 𝑀𝑋 (𝑡) = 𝐸(𝑒 𝑡𝑥 )
𝑛

= ∑ 𝑒 𝑡𝑥 𝑛𝐶𝑥 𝑝 𝑥 𝑞 𝑛−𝑥
𝑥=0
𝑛

= ∑ 𝑛𝐶𝑥 (𝑝𝑒 𝑡 )𝑥 𝑞 𝑛−𝑥 = (𝑞 + 𝑝𝑒 𝑡 )𝑛

𝑥=0

Characteristic Function of Binomial distribution:

∅𝑋 (𝑡) = 𝐸(𝑒 𝑖𝑡𝑥 )

𝑛

= ∑ 𝑒 𝑖𝑡𝑥 𝑛𝐶𝑥 𝑝 𝑥 𝑞 𝑛−𝑥

𝑥=0
𝑛

= ∑ 𝑛𝐶𝑥 (𝑝𝑒 𝑖𝑡 )𝑥 𝑞 𝑛−𝑥 = (𝑞 + 𝑝𝑒 𝑖𝑡 )𝑛

𝑥=0
Problem:

If the moment generating function of a random variable 𝑋 is of the form (0.4 𝑒 𝑡 + 0.6)8, find the
moment generating function of 3𝑋 + 2.

Solution:
Moment generating function of a random variable 𝑋 is

𝑀𝑋 (𝑡) = 𝐸(𝑒 𝑡𝑥 ) = (𝑞 + 𝑝𝑒 𝑡 )𝑛 = (0.6 + 0.4 𝑒 𝑡 )8

𝑋 follows the Binomial distribution with 𝑞 = 0.4, 𝑝 = 0.6, 𝑛 = 8

MGF of 3𝑋 + 2 is
𝑀3𝑋+2 (𝑡) = 𝐸(𝑒 𝑡(3𝑥+2) ) = 𝐸(𝑒 𝑡3𝑥 𝑒 2𝑡 ) = 𝑒 2𝑡 𝐸(𝑒 𝑡3𝑥 )
= 𝑒 2𝑡 𝐸(𝑒 (3𝑡)𝑥 )
= 𝑒 2𝑡 (0.4 𝑒 3𝑡 + 0.6)8

Cumulative Binomial distribution:

𝑥
𝑥
𝐵(𝑥; 𝑛, 𝑝) = 𝑃(𝑋 ≤ 𝑥) = ∑ 𝐵(𝑘; 𝑛, 𝑝) = ∑ 𝑛𝐶𝑘 𝑝𝑘 𝑞 𝑛−𝑘 , 𝑥 = 0,1,2,3, … , 𝑛
𝑘=0
𝑘=0

The Binomial probabilities can be obtained from cumulative distribution as follows

𝐵(𝑥; 𝑛, 𝑝) = 𝐵(𝑥; 𝑛, 𝑝) − 𝐵(𝑥 − 1; 𝑛, 𝑝)
Note: B(-1)=0
By using the Binomial table these can also be obtained.
Example:
The manufacture of large high-definition LCD panels is difficult, and a moderately high
proportion have too many defective pixels to pass inspection. If the probability is 0.3 that an
LCD panel will not pass inspection, what is the probability that 6 of 18 panels, randomly selected
from production will not pass inspection?
Solution:
X: LCD panel not pass in inspection.
n=18, p=0.30 and x=6
𝑏(𝑥; 𝑛, 𝑝) = 𝐵(𝑥; 𝑛, 𝑝) − 𝐵(𝑥 − 1; 𝑛, 𝑝)
𝑏(6; 18,0.30) = 𝐵(6; 18, 0.30) − 𝐵(5; 18, 0.30) = 0.7217 − 0.5344 = 0.1873
Poisson Distribution:

S.D. Poisson (1837) introduced Poisson distribution as a rare distribution of rare events.
i.e. The events whose probability of occurrence is very small but the no. of trials which could
lead to the occurrence of the event, are very large.
Ex:
1. The no. of printing mistakes per page in a large text
2. Number of suicides reported in a particular city
3. Number of air accidents in some unit time
4. Number of cars passing a crossing per minute during the busy hours of a day, etc.

Definition:
A random variable X taking on one of the non-negative values 0,1,2,3,4, … (i.e. which do not
have a natural upper bound) with parameter λ, λ >0, is said to follow Poisson distribution if its
probability mass function is given by

λ𝑥 𝑒 −λ
𝑃(𝑥; λ) = 𝑃(𝑋 = 𝑥) = { 𝑥! , 𝑥 = 0,1,2,3, …
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Then X is called the Poisson random variable and the distribution is known as Poisson
distribution.
And the Poisson parameter, λ = np>0

Conditions to follow in Poisson Distribution:

 The no. of trials ‘n’ is very large
 The probability of success ‘p’ is very small
 λ = np is finite.
Mean of Poisson distribution:
∞ ∞
λ𝑥 𝑒 −λ
𝜇 = 𝐸(𝑋) = ∑ 𝑥 𝑃(𝑥; λ) = ∑ 𝑥
𝑥!
𝑥=0 𝑥=0

𝜇 = 𝐸(𝑋) = λ = np

Variance of Poisson distribution:

Variance 𝑉(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋)2
∞

= ∑ 𝑥 2 𝑝(𝑥; 𝛌) − 𝜇 2
𝑥=0

𝑉(𝑋) = 𝜎 2 = λ

Cumulative Poisson distribution:

λ𝑘 𝑒 −λ
𝐹(𝑥; λ) = 𝑃(𝑋 ≤ 𝑥) = ∑𝑥𝑘=0 𝑃(𝑘; λ) = ∑𝑥𝑘=0 𝑘!

Moment generating function:

∞ ∞
𝑡𝑋 ) 𝑡𝑥 𝑡𝑥
λ𝑥 𝑒 −λ 𝑡
𝑀(𝑡) = 𝐸(𝑒 = ∑ 𝑒 . 𝑃(𝑥; λ) = ∑ 𝑒 = 𝑒 λ(𝑒 −1)
𝑥!
𝑥=0 𝑥=0
𝑡 −1)
𝑀(𝑡) = 𝑒 λ(𝑒
Characteristic function:
∞ ∞
𝑖𝑡𝑋 𝑖𝑡𝑥
λ𝑥 𝑒 −λ 𝑖𝑡
∅(𝑡) = 𝐸(𝑒 ) = ∑𝑒 . 𝑃(𝑥; λ) = ∑ 𝑒 𝑖𝑡𝑥 = 𝑒 λ(𝑒 −1)
𝑥!
𝑥=0 𝑥=0

𝑖𝑡 −1)
∅(𝑡) = 𝑒 λ(𝑒
Problem 1:

A hospital switch board receives an average of 4 emergency calls in a 10-minute interval. What
is the probability that
i. there at most 2 emergency calls in a 10-minute interval
ii. there are exactly 3 emergency calls in a 10-minute interval.

Solution:

Mean =λ = 4
𝑒 −λ λ𝑥
𝑃(𝑋 = 𝑥) = 𝑝(𝑥) =
𝑥!
i. P(at most 2 calls) = 𝑃(𝑋 ≤ 2)
= 𝑃(𝑋 = 0) + 𝑃(𝑋 = 1) + 𝑃(𝑋 = 2)
1 1 1
= + 4. + 8.
𝑒4 𝑒4 𝑒4
1
= (1 + 4 + 8) = 0.2381
𝑒4

1 16
ii. P(Exactly 3 calls) = 𝑃(𝑋 = 3) = 𝑒 4 . 3! = 0.1954

Problem 2:

If a random variable has a Poisson distribution such that P (1) =P (2). Find
i. Mean of the distribution
ii. P(4)
iii. 𝑃(𝑋 ≥ 1)
iv. 𝑃(1 < 𝑋 < 4)

Solution:
𝑒 −λ λ1 𝑒 −λ λ2
=
1! 2!
 λ2 = 2λ
λ = 0 or 2
But λ ≠ 0 or 2
Therefore λ = 2
i. Mean of the distribution is λ = 2
𝑒 −λ λ𝑥
ii. p(x) =
𝑥!

𝑒 −2 24
p(4) = = 0.09022
4!
iii. 𝑃(𝑋 ≥ 1) = 1 − 𝑃(𝑋 < 1) = 1 − 𝑃(𝑋 = 0)
𝑒 −2 20
=1− = 0.8647
0!
iv. 𝑃(1 < 𝑋 < 4) = 𝑃(𝑋 = 2) + 𝑃(𝑋 = 3)
𝑒 −2 22 𝑒 −2 23
= + = 0.4511
2! 3!
Problem 3:

Fit a Poisson distribution to the following data

𝑥 0 1 2 3 4 5
𝑓 142 156 69 27 5 1

Solution:

∑ 𝑓 𝑖 𝑥𝑖 0+156+138+81+20+5
Mean = ∑ 𝑓𝑖
= =1
400

Mean of the distribution is λ = 1

So, theoretical frequency for 𝑥 successes are given by 𝑁 𝑃(𝑥).
𝑒 −1 1𝑥
𝑁 𝑃(𝑥) = 400 × , 𝑥 = 0,2,3,4,5
𝑥!

i.e., 400 × 𝑒 −1 , 400 × 𝑒 −1 , 200 × 𝑒 −1 , 66.67 × 𝑒 −1 , 16.67 × 𝑒 −1 , 3.33 × 𝑒 −1

i.e., 147.15, 147.15, 73.58, 24.53, 6.13, 1.23
The expected frequencies are

𝑥 0 1 2 3 4 5
Theoretical 142 156 69 27 5 1
frequency
Expected 147 147 74 25 6 1
frequency

Problem 4:

𝑡 −1)
If the moment generating function of the random variable is 𝑒 4(𝑒 , 𝑓𝑖𝑛𝑑 𝑃(𝑋 = 𝜇 + 𝜎), where
𝜇 and 𝜎 2 are the mean and variance of the Poisson random variable X.

Solution:

𝑡 −1) 𝑡 −1)
𝑀(𝑡) = 𝑒 λ(𝑒 = 𝑒 4(𝑒
Mean=Variance= λ = 4
Standard deviation = √4 =2
𝑃(𝑋 = 𝜇 + 𝜎) = 𝑃(𝑋 = 4 + 2) = 𝑃(6)
𝑒 −4 46
𝑃(𝑋 = 𝑥) = 𝑃(𝑋 = 6) = = 0.1042
6!

Try yourself:

1. The distribution of typing mistakes committed by a typist is given below. Assuming the
distribution to be Poisson, find the expected frequencies

𝑥 0 1 2 3 4 5
 𝑓 42 33 14 6 4 1

Hypergeometric Distribution:

A discrete random variable X is said to follow the hypergeometric distribution with parameters
𝑁, 𝑀 𝑎𝑛𝑑 𝑛, if it assumes only non-negative values and its probability mass function is given by
(𝑀
𝑘
)(𝑁−𝑀
𝑛−𝑘
)
, 𝑘 = 0,1,2,3, … , min(𝑛, 𝑀)
𝑃(𝑋 = 𝑥) = ℎ(𝑘; 𝑁, 𝑀, 𝑛) = { (𝑁𝑛)
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Where 𝑁 is a positive integer, 𝑀 is a positive integer not exceeding 𝑁 and 𝑛 is a positive integer
that is at most 𝑁.

(Or)

𝑀𝐶𝑘 × 𝑁 − 𝑀𝐶𝑛−𝑘
𝑃(𝑋 = 𝑥) = ℎ(𝑘; 𝑁, 𝑀, 𝑛) =
𝑁𝐶𝑛

Mean and Variance of Hypergeometric Distribution:

𝑛𝑀 𝑀
Mean is 𝐸(𝑋) = 𝑛𝑝 = , 𝑤ℎ𝑒𝑟𝑒 𝑝 =
𝑁 𝑁
𝑛𝑀(𝑁−𝑀)(𝑁−𝑛)
Variance is 𝑣𝑎𝑟(𝑋) = 𝑛𝑝𝑞 =
N2 (𝑁−1)

Problem 1:

A batch of 10 rocker cover gaskets contains 4 defective gaskets. If we draw samples of size 3
without replacement, from the batch of 10, find the probability that a sample contains 2 defective
gaskets. Also find mean and variance.

Solution:
 𝑀𝐶𝑘 × 𝑁 − 𝑀𝐶𝑛−𝑘
𝑃(𝑋 = 𝑥) = ℎ(𝑘; 𝑁, 𝑀, 𝑛) =
𝑁𝐶𝑛
Here 𝑁 = 10, 𝑀 = 4, 𝑛 = 3, 𝑘 = 2
4𝐶2 × 6𝐶1
𝑃(𝑋 = 2) = ℎ(𝑘; 𝑁, 𝑀, 𝑛) = = 0.3
10𝐶3
𝑛𝑀 3×4 12
Mean is 𝐸(𝑋) = 𝑛𝑝 = = = = 1.2
𝑁 10 10
𝑛𝑀(𝑁−𝑀)(𝑁−𝑛) 3×4(10−4)(10−3)
Variance is 𝑣𝑎𝑟(𝑋) = 𝑛𝑝𝑞 = = = 0.56
N2 (𝑁−1) 102 (10−1)

Problem 2:

In the manufacture of car tyres, a particular production process is known to yield 10 tyres with
defective walls in every batch of 100 tyres produced. From a production batch of 100 tyres, a
sample of 4 is selected for testing to destruction. Find

i. the probability that the sample contains 1 defective tyre

ii. the expectation of the number of defectives in samples of size 4
iii. the variance of the number of defectives in samples of size 4.

Solution:

Sampling is clearly without replacement and we use the hypergeometric distribution with
𝑁 = 100, 𝑀 = 10, 𝑛 = 4, 𝑘 = 1
𝑀𝐶𝑘 × 𝑁−𝑀𝐶𝑛−𝑘
i. 𝑃(𝑋 = 𝑥) =
𝑁𝐶 𝑛

10𝐶1 × 100 − 10𝐶4−1 10 × 117480

𝑃(𝑋 = 𝑥) = = = 0.299 ≈ 0.3
100𝐶4 3921225
ii. The expectation of the number of defectives in samples of size 4
𝑛𝑀 4 × 10
𝐸(𝑋) = = = 0.4
𝑁 100
iii. The variance of the number of defectives in samples of size 4
𝑛𝑀(𝑁 − 𝑀)(𝑁 − 𝑛) 4 × 10(100 − 10)(100 − 4)
𝑣𝑎𝑟(𝑋) = = = 0.349
N2 (𝑁 − 1) 1002 (100 − 1)
Multinomial distribution:
 This distribution can be regarded as a generalization of Binomial distribution.
 When there are more than two mutually exclusive outcomes of a trial, the observations
lead to multinomial distribution. Suppose 𝐸1 , 𝐸2 ,…, 𝐸𝑘 are 𝑘 mutually exclusive
outcomes of a trial with respective probabilities.
 Th probability that 𝐸1 occurs 𝑥1 times, 𝐸2 occurs 𝑥2 times,… and 𝐸𝑘 , occurs 𝑥𝑘 times 𝑛
independent observations, is given by
𝑝(𝑥1 , 𝑥2 , … , 𝑥𝑘 ) = 𝑐𝑝1 𝑥1 𝑝2 𝑥2 … 𝑝𝑘 𝑥𝑘 , 𝑤ℎ𝑒𝑟𝑒 ∑ 𝑥𝑖 = 𝑛 𝑎𝑛𝑑 𝑐 is the number of
permutations of the events 𝐸1 , 𝐸2 , … , 𝐸𝑘 .
 To determine 𝑐, we have to find the number of permutations of 𝑛 objects of which 𝑥1 are
of one kind, 𝑥2 of another kind, …, 𝑥𝑘 of another kind 𝑘 𝑡ℎ kind, which is given by

𝑛!
𝑐=
𝑥1 ! 𝑥2 ! … 𝑥𝑘 !

𝑛! 𝑥1
Hence 𝑝(𝑥1 , 𝑥2 , … 𝑥𝑘 ) = 𝑝 𝑝2 𝑥2 … 𝑝𝑘 𝑥𝑘 , 0 ≤ 𝑥𝑖 ≤ 𝑛
𝑥1 !𝑥2 !…𝑥𝑘 ! 1

𝑛!
𝑝(𝑥1 , 𝑥2 , … 𝑥𝑘 ) = 𝑝 = ∏𝑘 ∏𝑘𝑖=1 𝑝𝑖 𝑥𝑖 (1)
𝑖=1 𝑥𝑖 !

Which is the required probability function of the multinomial distribution. Eq. (1) is called in
multinomial expansion
𝑘
𝑛
(𝑝1 + 𝑝2 + … + 𝑝𝑘 ) , ∑ 𝑝𝑖 = 1
𝑖=1

Since, the total probability is 1, we have

 𝑛!
∑ 𝑝(𝑥) = ∑ [ 𝑝 𝑥1 𝑝 𝑥2 … 𝑝𝑘 𝑥𝑘 ]
𝑥1 ! 𝑥2 ! … 𝑥𝑘 ! 1 2
𝑥 𝑥
𝑛
= (𝑝1 + 𝑝2 + … + 𝑝𝑘 ) = 1
Moments of multinomial distribution:
𝑘
𝑀(𝑡) = 𝑀𝑋1 , 𝑋2 ,… ,𝑋𝑘 (𝑡1 , 𝑡2 , … , 𝑡𝑘 ) = 𝐸 (𝑒 ∑𝑖=1 𝑡𝑖𝑋𝑖 )
𝑛
= (𝑝1 𝑒𝑡1 + 𝑝2 𝑒𝑡2 + … + 𝑝𝑘 𝑒𝑡𝑘 )
Moments:
𝐸(𝑋𝑖 ) = 𝑛𝑝𝑖
𝑉𝑎𝑟(𝑋𝑖 ) = 𝑛𝑝𝑖 (1 − 𝑝𝑖 )

Problem 1:

Suppose we have a bowl with 10 marbles 2 red marbles, 3 green marbles, and 5 blue marbles.
We randomly select 4 marbles from the bowl, with replacement. What is the probability of
selecting 2 green marbles and 2 blue marbles?

Solution:

To solve this problem, we apply the multinomial formula. We know the following:

The experiment consists of 4 trials, i.e., 𝑛 = 4.

4 trials produce 0 red marbles, 2 green marbles, and 2 blue marbles;

i.e., 𝑥𝑟𝑒𝑑 = 𝑥1 = 0 , 𝑥𝑔𝑟𝑒𝑒𝑛 = 𝑥2 = 2 , and 𝑥𝑏𝑙𝑢𝑒 = 𝑥3 = 2

On any particular trial, the probability of drawing a red, green, and blue marble is 0.2, 0.3, and
0.5, respectively.

i.e., 𝑝𝑟𝑒𝑑 = 0.2, 𝑝𝑔𝑟𝑒𝑒𝑛 = 0.3, 𝑝𝑏𝑙𝑢𝑒 = 0.5

The multinomial formula is

 𝑘
𝑛!
𝑝(𝑥1 , 𝑥2 , … 𝑥𝑘 ) = ∏ 𝑝𝑖 𝑥𝑖
∏𝑘𝑖=1 𝑥𝑖 !
𝑖=1

𝑛! 𝑥1
i.e., 𝑝 = 𝑝 𝑝2 𝑥2 𝑝3 𝑥3
𝑥1 !𝑥2 !𝑥3 ! 1

4!
= (0.2)0 (0.3)2 (0.5)2 = 0.135
0! 2! 2!

𝑝 = 0.135

Thus, if we draw 4 marbles with replacement from the bowl, the probability of drawing 0 red
marbles, 2 green marbles, and 2 blue marbles is 0.135.

Problem 2:

In India, 30% of the population has a blood type of O+, 33% has A+, 12% has B+, 6% has AB+,
7% has O-, 8% has A-, 3% has B-, and 1% has AB-. If 15 Indian citizens are chosen at random,
what is the probability that 3 have a blood type of O+, 2 have A+, 3 have B+, 2 have AB+, 1 has
O-, 2 have A-, 1 has B-, and 1 has AB-?

Solution:

𝑛 = 15 trials

𝑝1 = 30% = 0.30 (probability of O+)

𝑝2 = 33% = 0.33 (probability of A+)

𝑝3 = 12% = 0.12 (probability of B+)

𝑝4 = 6% = 0.06 (probability of AB+)

𝑝5 = 7% = 0.07 (probability of O−)

𝑝6 = 8% = 0.08 (probability of A−)

𝑝7 = 3% = 0.03 (probability of B−)

 𝑝8 = 1% = 0.01 (probability of AB−)

𝑥1 = 3 (3 O+)

𝑥2 = 2 (2 A+)

𝑥3 = 3 (3 B+)

𝑥4 = 2 (2 AB+)

𝑥5 = 1 (1 O−)

𝑥6 = 2 (2 A−)

𝑥7 = 1 (1 B−)

𝑥8 = 1 (1 𝐴B−)

𝑘 = 8 (8 𝑝𝑜𝑠𝑠𝑖𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠)

𝑛!
𝑝= 𝑝 𝑥1 𝑝 𝑥2 𝑝 𝑥3 𝑝 𝑥4 𝑝 𝑥5 𝑝 𝑥6 𝑝 𝑥7 𝑝 𝑥8
𝑥1 ! 𝑥2 ! 𝑥3 ! 𝑥4 ! 𝑥5 ! 𝑥6 ! 𝑥7 ! 𝑥8 ! 1 2 3 4 5 6 7 8

15!
= × 0.303 × 0.332 × 0.123 × 0.062 × 0.071 × 0.082 × 0.031
3! 2! 3! 2! 1! 2! 1! 1!
× 0.011

𝑝 = 0.000011

Discrete Bivariate Distributions:

Covariance:

We are often interested in the inter-relationship, or association, between two random variables.
The covariance of two random variables X and Y is
𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸(𝑋𝑌) − 𝐸(𝑋)𝐸(𝑌)
Or
𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸[(𝑋 − 𝐸(𝑋))(𝑌 − 𝐸(𝑌))]

Note:

Covariance is an obvious extension of variance

𝐶𝑜𝑣(𝑋, 𝑋) = 𝑉𝑎𝑟(𝑋)

Moments:

1. The moments for the Binomial Distribution:

By the definition of a moment 𝜇𝑟 = 𝐸[𝑋 − 𝐸(𝑋)]𝑟
The first four central moments of the Binomial distribution are:

we know that 𝜇0 = 1, 𝜇1 = 0

𝑀𝑒𝑎𝑛 = 𝑛𝑝
𝜇2 = 𝑛𝑝𝑞
𝜇3 = 𝑛𝑝𝑞(𝑞 − 𝑝)
𝜇4 = 𝑛𝑝𝑞[1 + 3𝑝𝑞(𝑛 − 2)]

2. The moments for the Poisson Distribution:

The first four central moments of the Poisson distribution are:

𝜇1 = 0
𝜇2 = 𝜆
𝜇3 = 𝜆
𝜇4 = 3𝜆2 + 𝜆