SUCA 203

UNDERGRADUATE COURSE
BCA - COMPUTER APPLICATIONS

SECOND YEAR
THIRD SEMESTER

CORE PAPER - VII

NUMERICAL AND
STATISTICAL METHODS

INSTITUTE OF DISTANCE EDUCATION

UNIVERSITY OF MADRAS
BCA, COMPUTER APPLICATIONS CORE PAPER - VII
SECOND YEAR NUMERICAL AND STATISTICAL
THIRD SEMESTER METHODS

WELCOME
Warm Greetings.

It is with a great pleasure to welcome you as a student of Institute of Distance

Education, University of Madras. It is a proud moment for the Institute of Distance education
as you are entering into a cafeteria system of learning process as envisaged by the University
Grants Commission. Yes, we have framed and introduced Choice Based Credit
System(CBCS) in Semester pattern from the academic year 2018-19. You are free to
choose courses, as per the Regulations, to attain the target of total number of credits set
for each course and also each degree programme. What is a credit? To earn one credit in
a semester you have to spend 30 hours of learning process. Each course has a weightage
in terms of credits. Credits are assigned by taking into account of its level of subject content.
For instance, if one particular course or paper has 4 credits then you have to spend 120
hours of self-learning in a semester. You are advised to plan the strategy to devote hours of
self-study in the learning process. You will be assessed periodically by means of tests,
assignments and quizzes either in class room or laboratory or field work. In the case of PG
(UG), Continuous Internal Assessment for 20(25) percentage and End Semester University
Examination for 80 (75) percentage of the maximum score for a course / paper. The theory
paper in the end semester examination will bring out your various skills: namely basic
knowledge about subject, memory recall, application, analysis, comprehension and
descriptive writing. We will always have in mind while training you in conducting experiments,
analyzing the performance during laboratory work, and observing the outcomes to bring
out the truth from the experiment, and we measure these skills in the end semester
examination. You will be guided by well experienced faculty.

I invite you to join the CBCS in Semester System to gain rich knowledge leisurely at
your will and wish. Choose the right courses at right times so as to erect your flag of
success. We always encourage and enlighten to excel and empower. We are the cross
bearers to make you a torch bearer to have a bright future.

With best wishes from mind and heart,

DIRECTOR

(i)
BCA, COMPUTER APPLICATIONS CORE PAPER - VII
SECOND YEAR NUMERICAL AND STATISTICAL
THIRD SEMESTER METHODS

COURSE WRITERS
Mrs. V. Parimala
Guest Lecturer
Quaidae Millath Arts and Science College
Chennai 600 002.
(Lessons 1 to 14)

Mr. K. Rajasekaran
Lecturer in Computer Science
D.B. Jain College
Chennai - 600 096.
(Lessons 15 to 26)

EDITING AND COORDINATION

Dr. S. Sasikala
Assistant Professor in Computer Science
Institute of Distance Education
University of Madras
Chepauk Chennnai - 600 005.

ISBN No: 978-81-19111-65-7

Year of Publication: 2018

Total No. of Pages: 330

Printed by: Publication Section, Institute of Distance Education IIDE), University of Madras,
Chennai - 600 005, Website:- [Link]

(ii)
 BACHELOR OF COMPUTER APPLICATIONS
SECOND YEAR
THIRD SEMESTER
CORE PAPER - VII
NUMERICAL AND STATISTICAL METHODS
SYLLABUS
Objective of the course

This course introduces the concepts of Numerical Analysis and Statistical Methods

Course outline

Unit-1: Introduction- Mathematical Preliminaries- Errors: Computations, Formula -

Errors in a Series Approximation- Roots of Equations- Linear Equations: Bisection , False
Position Methods- Newton-Raphson Method- Secant Method- Muller’s Method- Lin-
Bairstow’s Method- Simultaneous Linear Equations: Matrix Inversion Method- Gauss
Elimination, Gauss-Jordan, LU Decomposition Methods- Gauss-Seidel Method.

Unit-2: Numerical Differentiation- Errors in Numerical Differentiation- Cubic Spline

Method- Numerical Integration- Trapezoidal Rule- Simpson’s 1/3 and 3/8 Rules- Romberg
Integration- Ordinary Differential Equations- Taylor’s Series Method- Euler’s Method- Runge-
Kutta 2nd and 4th Order Methods-Predictor-Corrector Methods.

Unit-3: Sampling- Frequency Distribution- Cumulative Frequency Function- Grouped

Sample- Measures of Central Tendency: Mean, Median and Mode- Geometric Mean-
Harmonic Mean – Dispersion: Range, Mean Deviation, Variance and Standard Deviation-
Moments- Computation of Moments

Unit-4:Probability- Characteristics: Addition, Multiplication and Conditional Probability

Laws- Discrete Distributions: Random Variable- Density and Distribution Functions.- Binomial
Distribution- Poisson Distribution- Hypergeometric Distribution- Mathematical Expectation.

Unit-5: Correlation and Regression Analysis: Linear Least Squares Fit- Nonlinear
Fit- Fitting a Polynomial Function- Coefficient of Correlation- Properties- Multiple Correlation

(iii)
– Partial Correlation- Rank Correlation- Tests of Significance- Chi square Test- Goodness
of Fit, Algorithm and Analysis of Contingency Tables- t-Test and F-Test.

1. Recommended Texts:

i) [Link], 2005,Introductory Methods of Numerical Analysis, 4th Edition, Prentice-

Hall of India Pvt. Ltd..

ii). [Link] , 2000, Computer Oriented Statistical and Numerical Methods-

Macmillan India Ltd.

2. Reference Books:

i) V. Rajaraman,2005, Computer Oriented Numerical Methods, 3rd Edition, Prentice-

Hall of India Pvt. Ltd..

ii) K. S. Trivedi,2005,Probability and Statistics with Reliability, Queuing and Computer

Science Applications, Prentice-Hall of India Pvt. Ltd.

iii) E. Balagurusamy,1999, Numerical Methods, Tata McGraw-Hill Publishing Co. Ltd..

iv) P. Niyogi,2003,Numerical Analysis and Algorithms, Tata McGraw-Hill Publishing Co.

Ltd..

(iv)
 BACHELOR OF COMPUTER APPLICATIONS
SECOND YEAR
THIRD SEMESTER
CORE PAPER - VII
NUMERICAL AND STATISTICAL METHODS

SCHEME OF LESSONS

[Link]. Title Page No.

1. Roots of Equations 001

2. Solution of Simulataneous Equations 035

3. Numerical Integration 067

4. Differential Equations 080

5. Diagrammatic and Graphical Representations 095

6. Measures of Central Tendencies 109

7. Measures of Dispersion 124

8. Theory of Probability 152

9. Mathematical Expectation 177

10. Discrete Distributions 195

11. Continuous Distributions 226

12. Correlation and Regression 256

13. Sampling and Large Sample Tests 291

14. Small Sample Tests 305

(v)
 1

LESSON - 1
ROOTS OF EQUATIONS
Structure

1.1 Introduction

1.2 Learning Objectives

1.3 Graphical Method

1.4 Bisection Method

1.5 Regular Falsi Method

1.6 Fixed - Point Iteration

1.7 Newton - Raphson

1.8 Secant Method

1.9 Muller's Method

1.10 Bairstow's Method

1.11 Summary

1.12 Check Your Answers

1.13 Model Questions

1.1 Introduction
In general, an equation in x may be written as f(x) = 0. The root of the equation are those
values of x for which the equation is satatisfied. The equations are two types. (1) polynomial
equation (2) Transcendental equation.

anxn + an-1 xn-1+ ............a0 = 0 where a, # 0. This is called the nth degree polynomial and
has n roots.
2

These roots may be:

1. Real and different,

2. Real and equal,

3. Real and Complex. Complex roots appear in pairs of the form a +iß, a-iß

f(x) = 2ex. sinx-x. This equation is called a transcendental equation. The zero of f(x) give
the roots of the equation. The roots can be found either by graphically or by iterative methods
such as the Newton Raphson method.

1.2 Learning Objectives

After studying the lesson, you should be able to understand

• The process of finding the roots of f(x) = 0.

• To understand to calculate the roots of the equation by the methods of bisection, falsi,
Newton Rephson & secant,

• To calculate the real and complex roots of polynomial using Muller's method and
Bairstow's method.

1.3 Graphical Methods

A simple method for obtaining an estimate of the root of the equation f(x) = 0 is to make
a plot of the function f(x) and observe where it crosses the x axis. This point which represents
the x value for which f(x) = 0 provides an approximation of the root.

Problem 1

Use the graphical method to determine a root of the equation.

f(x) = 667.38 / x (1-e-0.146543x) - 40

Solution

From the following table for various values of x

 3

Using these values, plot the curve. The resulting curve crosses the x axis between 12 and 16
visual inspection of the plot provides a rough estimate of the root 14.75. Substitute the approximate
value in f(x)

f(14.75) = 0.059 (approx)

which is close to zero.

1.4 Interval Bisection Method

This method is also known as Balzano bisection method named after the German
mathematician Bernhard Balzona who invented this method. This is one of the simplest method,
but the disadvantage is its slow convergence to the root.

Let y = f(x) be a continuous function in [a,b] and let f(a) be positive and f(b) be negative.
Then there is root of f(x) = 0 between a and b.

Let its approximate value be x0 = a + b/2

4

If f(x) = 0 it imply that x, is a root of f(x) = 0. If f(x0) is negative the root of f(x) = 0 lies
between a and x0 and if f(x0) > 0 the root lies between x0 and b. Choose that interval in which the
root lies and bisect it as before. Continue this process until a root is obtained to the desired of
accuracy. The method is illustrated in the Fig. 1.1.

Fig. 1.1 Roots of the equation

In the fig 1.2, the first approximation

x0 = a+b/2 and f(x0) <0.

Hence the root lies between a and x, the next approximation is x, = X? + a / 2 and f(x,)>0.

Hence the root lies between Xo and X,

We take x2 = Xo + X/2

as the third approximation and so on. The bisection method is simple but slowly
convergent method.
 5

Problem 2

Obtain a root of x3-x-1= 0. Correct to three decimal places using the bisection method.

Solution

(1) = 1-2-1 <0

f(2) = 23-35>0

There is a root of f(x) = 0 between 1 and 2

Take x0 = 1+2 / 2 = 1.5 to be the first approximation of the root.

f(1.5) = (1.5) - 1.5-1.

= 3.375-2.5= 0.875

Since f(1) < 0 and f(1.5)>0 the root of f(x) = 0 lies between 1 and 1.5.

Take x1 = 1 + 1.5 /2 = 1.25 as second approximation

f(1.25) (1.5)3 -1.25 -1

= 1.9531-2.25 = 0.719

The root of f(x) lies between 1.25 and 1.50

Let x3 = 1.25 +1.50 / 2 = = 1.375

f(1.375) = (1.375)3 - 1.375 - 1

= 2.5996-2.375

= 0.2246

The root of f(x) = 0 lies between 1:25 and 1.375

Let x4 = 1.25 +1.375 / 2 = 1.3125

f(1.3125) = (1.3125)3 - 1.3125-1

= 2.2610-2.3125

= 0.0515

The root of f(x) = 0 lies between 1.375 and 1.3125

6

Let x5 = 1.375 +1.3125 / 2 = 1.3438

f(1.3438) = (1.3438)3-1.3438-1

= 2.4266-2.3438

= 0828

The root of f(x) = 0 lies between 1.3125 and 1.3438

Let x6 = 1.325+1.3438 / 2

f(1.3282) = (1.3282)3-1.3282-1

= 2.3431-2.3282

= 0.0149

Let x7= 1.3125 +1.3282 / 2 = 1.3204

f(1.3204) = (1.3204)2-1.3204-1

= 2.3021-2.3204

= 0.0183

The root of f(x) = 0 lies between 1.3282 and 1.3204

Let x8= 1.3282+1.3204 / 2 = 1.3243

f(1.3243) = (1.3243)3-2.3243

= 2.3225-2.3243

= 0.0018

The root of f(x) = 0 lies between 1.3282 and 1.3243

Let x9 = 1.3282+1.3243 / 2 = 1.3263

= 2.3331 -2.3263 > 0

The root of f(x) = 0 lies between 1.3243 and 1.3263

X10 = 1.3243 +1.3263 / 2 = 1.3253

f(1.3253) = (1.3253)3-2.3253
 7

= 2.3278-2.3253

= 0.0025

The root of f(x) = 0 lies between 1.3243 and 1.3253

Let x11 = 1.3243 +1.3253 / 2 = 1.3248

f(1.3248) = (1.3248)3-2.3248

= 2.3251-2.3248788621)

= 0.0003

Since f(1.3248) is approaching zero, 1.3248 is a good approximation to the root.

1.5 Regular Falsi Method or False Position Method

Another popular method of solving equations is the false position or regula falsi method. Let f(x)
be a continuous function in [a,b]. Let f(a) > 0 and f(b) < 0. Connect two points A(a), f(a)) and B(b,
f(b)) by a straight line L. Let the line cut x axis at c (x, 0) as in Fig 1.2. To find X., we find the slope
of AB and BC and equate [Link].

Fig. 1.2
8

Slope of AB is f(b) - f(a) / b- a

Slope of BC is f(b) - 0 / b - x0

We get f(b) - f(a) / b - a = f(b) / b - x0

(b-x0) [f(b) - f(a)] = (b-a) f(b)

(b - a)f(b)
(b-x0) = f(b) - f(a)

x0 = b- (b - a)f(b) / f(b) - f(a)

= b[f(b)-f(a)] (b - a)f(b)
f(b) - f(a)

= af (b) - bf(a)
f(b) - f(a)

Three possibilities arise. If f(x) = 0, then x, is a root of f(x) = 0 or if f(x) and f(a) are of
opposite signs, there is a root of f(x) = 0 in (a, x) or if f(x) and f(b) are of opposite signs, there is
a root of f(x) = 0 in (x01, b).

If the process is continued, we get a sequences of nested intervals { [a, b]} each of
which contain a root. The approximation to the root in the nth iteration is

f(bn)(bn-an)
x=b2 - f(bn) - f(a)

Problem 3

Find a real root of the equation f(x) = x3-2x-5=0

Solution

f(2)=8-4-5=-1 <0

f(3) = 27 - 6 - 5 = 16 > 0

There is a root of f(x) = 0 between 2 and 3.

 9

Applying the formula

f(b1)(b1 - a1)]
x1 = b 1 - f(b1) - f(a1)

We get x1 = 2- (2-3)(-1)
-1-16

=2+- 1 / 17 = = 2.059

f(x1) = f(2.059) = (2.059)3 -2(2.059) - 5

= 8.7291-9.1180 -0.3889.

Hence there is a root of f(x) = 0 in (2.059, 3)

As a second approximation, we have

(2.059 3)(-0.3889)
X2 = (2.059) -
-0.3889 - 16

(0.9413) x (-0.3889)
= (2.059) +
16.3889

= 2.059 +0.0223 = 2.0813

f(x2) = f(2.0813) = (2.0813)-2(2.0813)-5

= 9.0518-9.1626

= -0.1468

There is a root of f(x) = 0 between 2.0813 and 3.

(2.08133)(-0.1468)
X3 = (2.0813) -
(-0.1468) - 16

(0.9187)(0.1468)
= (2.0813) +
16.1468

= 2.0813+0.0084

= 2.0897
10

f(x3) = (2.0897)3-2(2.0897) - 5

= 9.1254-9.1974

= -0.054

The root is lies between 2.0897 and 3

(2.0897-3) (-0.054)
x4 = 2.0897-
(-0.054)-16

0.9103 x 0.054
= 02.0897 +
16.054

= 2.0897 + 0.0031 = 2.0928

f(2.0897) = (2.0897)3-2(2.0928) - 5

= 9.1661-9.1856

= -0.0195 <0

The roots lies between 2.0928 and 3.

99s wen s printsido to un brs (x=xr "x Jeg sw *x lo

X5 = 2.0928- (2.09283) x 0.0195

0.0195 - 16

= 2.0928 + 0.9072 x 0.0195

16.0195

= 2.0928+0.0011

= 2.0939

f(2.0939) = (2.0939) - 2 (2.0939) - 5

= 9.1805-9.1878

= -.0073
 11

The root lies between 2.0939 and 3

(2.0939 - 3) x -0.0073
= 2.0939-
-0.0073 - 16

(0.9061) x (0.0073)
= 2.0939 +
16.073

= 2.0939+.0004

= 2.0943

The approximate of the root may be taken as 2.094.

1.6 Picard's Fixed Point Iteration

Definition: Let f(x) = 0 be written as x = F(x) where F(.) is a new function of x. If x is a root
of F(x), then it is called a fixed point of the function F.

Let f(x) = x2+5ex - 10. Then the problem of finding a root of the equation f(x) = 0 is
equivalent to each one of the following fixed point problems.

X= F1(x) = x2 + 5ex - 10 + x

X= F2(x) = 10 - 5ex

and X = F3(x) = In (10 - x2) / 5

Let x* be an exact solution of the equation f(x) = 0. Then f(x*) = 0 and x* = F(x*). Now the
successive approximation method consists of substituting the approximated value of x* in x =
F(x) and thus of obtaining a new approximation of x*. So with the initial estimate x(0) of x*, we
get

X(1) = F(x(0))

Similarly, we get

x(2) = F(x(1)), x(3) = F(x(2))

and in general

X(n) = F(x(n-1))
12

Now we have generated the sequence {x(n)}. Next, the problem is to know when this
sequence {x(n)} converages to the exact solution x* of x = F(x).

Termination of the Iteration

We learn from theoretic aspect that the solutions obtained only after infinite number of
iterations provided the method converages. In practice, we stop the iteration when one of the
following conditions is satisfied.

1. If f(x(n) is nearly zero ie |f(x("))| < ?,

2. If the improvement of x(n) from one iteration to the next iteration is not justified when
compared to the remaining calculations
x(n) - x(n-1)
(i.e.) < 2 or |x(n) - x(n-1)| 2
x(n)

3. If the convergence is not obtained before a certain given number of iterations, M. (i.e.)
n> M

4. In the iteration process, if the precision e, (or e2) can not be attained in a reasonable
number of iterations than process is considered as divergent for the given initial
approximation x(0).

Some times it is difficult to transform the problem f(x) = 0 into the form x = f(x) by simple
algebraic transformation. In these cases one can write

f(x) = 0 as x = x +  f(x) = F(x)

where e is a non-zero constant and is arbitrary. We choose e so that the derivative of

F(x) = x + 0 f(x) satisfies the condition

|F' (x)| = |1 +  f'(x) | < 1

in order to ensure the convergence.

Problem 4

Use Picard's iteration principle or fixed point iteration to locate the root of f(x) = ex-X.
 13

Solution

Note that f(x) = ex-x=0 can be written in the form x = F(x) as

x = e-x

In the iteration form x(n) = e-x(n-1)

Starting with intitial guess x(0) = 0, the above iterative equatiion can be applied to get the
following table

i X(1)  a% t%

0 0 - 100.00

1 1.00000 100.0 76.0

2 0.367879 171.8 35.1

3 0.692201 46.9 22.1

4 0.500473 38.3 11.8

5 0.606244 17.4 6.89

6 0.545396 11.2 3.83

7 0.579612 5.90 2.2

8 0.560115 3.48 1.24

9 0.571143 1.93 0.705

10 0.564879 1.11 0.399

Each Iteration brings the estimate closer to the true value 0.56714329

Problem 5

Find a real root of the equation x3 + x2-1=0 Picard's fixed point method.

Solution

Putting x = 0, we get f(0) = -1 <0

and x = 1, we get f(1) = 1 > 0

14

Hence a real root lies between 0 and 1. The given equation x3 - x2 - 1 = 0 can be written
as: x2(x + 1) -1 =0
 15

Hence the root is 0.7549 with four decimal accuracy.

Note: One can stop the iteration depending upon the some pre-specified error.

1.7 Newton - Raphson Method

Let p be a root of f(x) (i.e.) f(p) = 0. Assume that f(x), f'(x), f'(x) are continuous near p Now
we will construct a sequence (pk) that goes to p faster than bisection or false position method.
The Newton - Raphson method relies one the continuity of f' and f"

Assume that the initial approximation p, is near the root p. Then the Fig 1.3 y = f(x
intersects the x-axis at the point (p,0) and the point (p, f(p) lies on the curve near the poin (p, 0).

Define P1 to be the point of intersection of the line tangent to the curve at the point (P ff(p0)) and
the x-axis.

Now P1 is closer to p than Po

We Now slope of the line L is

0 - f(po)
m= (1)
P1 - Po
16

Slope of f at (Po, f(P0) is

m = f'(po) (2)

Therefore (1) & (2) gives

f(po)
p1 =
Po- f'(Po)
Repeat this process to obtain a sequence (pk) that converges to p. In general
f(Pn)
Pn+1 = Pn -
f'(Pn)

Look at Taylor's polynomial of deg 1 with remainder term

(x - Po)2
f(x) = f(p0) + f'(po) (x - p0) + f'(c)
2!

where p < c < P0. Put x = p. Then

(p - xo)2
f(p) = 0 = f(p0) + f'(po) (p - p0) + f'(c)
2!

If p0 is close to p, the last term in the above expression will be small compared to the
same of the other two terms and hence negligible.

Thus we have
f(po)
P~ Po - = P1 (say) p, is taken as an approximation to p and the iteration continues.
f'(Po)

Problem 6

Find a real root of x3-3x+1=0 which lies between 1 and 2

f(1) = 1-3+1 = -1 <0

f(2) = 8-6+1=3>0

f'(x) = 3x2 - 3
 17

Taking the first approximate root as 1 we get supe evit to loon art b
f(1) f'(1)
h=- - = 
(-1) 0

We shall take the approximation root to be 2

-f(2) 3
then h = =- = - 0.33
f'(2) 9

the root is 2-0.33 1.67

Taking the root as 1.67 we get

-f(1.67) (1.67)3 - 3(1.67) + 1
h= =
f'(1.67) 3(1.67)2 - 3

(4.657-5.01+1)
=
8.367-3

= 0.121

Next approximate to the root is

1.67 -0.121 = 1.549

f(1.549)
h=
f'(1.549)

[(1.549)3 - 3(1.549) + 1]
=-
3(1.549)2 - 3

[-3.7167 - 4.647 +1]

=-
4.1982

0.0697
=-
4.1982

= -0.0166

Next approximate root is

1.549 -0.0166 = 1.5324

18

Problem 7

Find the root of the equation ex-4x= 0, correct to three decimal places by Newton- Raphson
method.

f(x) = ex-4x

f(2) = e2-8

= 7.3891-8

= -0.6109<0

f(3) = e3-12

= 20.0855 12

= 8.0855 > n

There is a root of f(x) = 0 between 2 and 3

The root is near to x = 2

f(2)
h=
f'(2)

-(e2 -8)
=
(e2-4)

(7.3891-8)
=
(7.3891-4)
0.6109
=
3 .3891

= 0.1803

Approximate root is

2+0.1803 = 2.1803
19
20

1.8 The Secant Method

The formula involved in the secant method is the same one that is used in regua falsi
method except that the logical decisions regarding how to define each succeeding terms are
different. Then Fig 1.4 the two initial points (a,, f(a), (a,, f(a,) near the point (r,0) are needed.
Define a, to the point of intersection of the line through these two points.

Now a, will be closer to r than to either a, or a,. The equation relating a,, a2, a is found by
considering the slope way to lead met yd mere no leone
f(a1) - f(a0)
m =
a1 - a0

0 - f (a1)
=
a2 - a1
 21

the values in (*) are the slope of the scant line through the first two approximations the sbpe of
the through (a,, f(a1)), (a2, 0) respectively

Solving for a2 = g (a1, a0) and we get

Apparently, the secant method locates quite rapidly a point at which |f(x) is “small” but
gives, in general, no feeling for how far away from a zero of f(x) this point might be. Also (a) and
f(a) need not be found opposite sign, so that the expression (2) is prone to round off error
effects. In an extreme situation, we may have f(a) = f(a,), making the calculation of a., impossible.
It is better to calculate an from (1), we have the first divided difference which is the slope of the
chord (secant).

Comparison between the Secant method and False position method

Recall the formula for the secant method

and the formula for False position method

The are identical on a term by term basis. Both use two initial estimates to compute a
new estimate of the root. However, an important difference between the methods is how one of
22

the initial values is replaced by the new estimate. Recall that inthe false-position method, the
latest estimate of the root replaces whichever of the original values yielded a root. Therefore,
the false position method converges because the root is kept within the bracket. On the contrary,
the secant method replaces the values in strict sequence. With the value a,,, replacing a, and a,
replacing a,,. As a result the two values can sometimes lie on the same side of the root. For
certain cases, this can lead to divergence.

Although the secant method may be divergent, when it converges it usually does so at a
quicker rate than the false position method. In the false-position method, one end of the interval
stays fixed to maintain the bracketing of the root. This property prevents divergence but it is a
short coming with regard to the rate of convergence.

Problem 9

Use the secant method to find a root of f(x) = e-x -x with initial estimates of a0 = 0 and =
a1 = 1.0 Perform there iterations.

Solution

f(a1)(a?-a1) f(a.)- f(a,)

 23

After three iterations, we get the root as 0.56717

Problem 10

Evaluate a real root of 4sin x = e between 0 and 0.5 using the secant method. Perform
four iterations.

Solution

Now take f(x) = 4sin x-ex, where x is in terms of radians

a0=0, f(a0 = -1

a1 = 0.5, f(a1) = -0.08683363

a2 = 0.3650302,
an f(an)(an-an-1)
Using the formula an+1=
f(an)-f(an+1)

Similarly a3 = 0.369706

a4 = 0.370426

a5 = 0.3705376

So the value of the real root is 0.3705376

Note: Students are advised to work out the details here.3.0-

Problem 11

Find a root of 2sinh x-coshx = 0 between 0 and 1 using the secant method. Perform
three iterations.

Solution

Take a = 0, a, = 1, using the formula

Similarly a2 = 0.5533049

a3 = 0.549499

a4 = 0.5493155
24

Therefore an approximate real root is 0.549 with three decimal places accuracy.

1.9 Muller's Method

In the secant method, we obtain a root estimate by projecting a line to the x axis. through
two function values. In Muller's method, a root estimate is obtained by projecting a parabola
through three points.

The method consists of deriving the coefficients of the parabola which passess through
the three points. These coefficients can then be substituted into the quadratic formula to get the
point where the parabola intercepts the x-axis.

Write the parabolic equations in a convenient form

 25

To find the root of (*), we use an alternative formulation to find roots of quadratic equation
ax2 + b + c = 0. (i.e.)

In out case, with our notions,

Notice that the use of quadratic formula means that both real and complex roots can be
located. This is really a crowning achievement.

Error estimate

In (10), we get two roots because of ± is the denominator. In Muller's method, the sign is
chosen to agree with the sign of b which will give the root estimate closes to x2

Once x in (10) is determined, call it x3. Now we have to discard one point. This can be
done as follows.

1. The only real roots are being located, we choose the two original points that are nearest
to the new root estimate x3.

2. If both real and complex roots are being evaluated, a sequential approach is employed.
This is just like the secant method, x1, x2 and x3 take the place of x0, x1 and x2.
26

Problem 12

Use Muller's method with x0 = 4.5, x1 = 5.5 and x2 = 5 to determine a root of the equation
f(x) = x -13x2 - 12.
3

Solution

First evaluate the function at initial values

f(x0) = f(4.5)= 20.625

f(x1) = f(5.5) = 82.875

f(x2) = f(5.0) = 48.000

Now

h0 = x1-x0= 5.5 - 4.5 = 1

h1 = x2-x1 = 5.0 - 5.5 = -0.5

Using the above values, we get

Therefore,

a positive sign is employed in the denominator of

 27

Because the error is large, iterate with the following values for x0, x1 x2.

The new values are x0 = 5.5, x1 = 5, x2 = 3.976487.

Continuing the calculation, we get the following table

n xn Ea

0 5 -

1 3.976487 25.74

2 4.00105 0.6139

3 4 0.060

4 4 0.0000119

After four iteration, we get a root as 4.

Polynomial Deflation

Assume that we have determined as single root of an n-th order polynomial. If we repeat
the root location procedure, we might find the same root. So it is essential to remove the found
root before proceeding. This removal processes is referred to as polynomial deflation.

Problem 13 (Synthetic division)

Let

f(x) = x5 - 7x4 - 3x3 + 79x2 - 46x - 120

f(-1)=-1-1 +3 + 79 +46 - 120

= 128 128=0.
28

Therefore -1 is a root of f(x). Now we divide f(x) by x+1. This is done by synthetic division.

-1 -7 -3 79 -46 -120

0 -1 8 -5 -74 +120

1 -8 5 74 -120 0=R

Therefore the quotient polynomial is

q(x) = 1x4 - 8x3 + 5x2 + 74x-120

(i.e.) f(x) = (x + 1) q(x) + R

Where R = 0

Now we also find f(-3)= 0. We will find find the quotient after dividing f(x) by x+3

-1 -7 -3 79 -46 -120

0 -3 30 -81 6 120

1 -10 27 -2 -40 0=R

Here q(x) = x^ - 103 + 27x2 - 2x-40 and remainder is zero.

1.10 Bairstow's Method

Let fn(x) = a0 + a1 x + a2x2 + ...... + anxn

Bairstow's method divides the polynomial f(x) by a quadratic factors x2-rx-s. After division,
we will get

fn-2(x) = b2 + b3x + ... + bn-1xn-3 + bnxn-2

with a remainder R = b1 (x-r) + bo

As with normal synthetic division, a simple recurrence relationship can be used to perform
the division by the quadratic factor

bn = an (1)
 29

bn-1 = an-1 + rbn

b1 = a1 + rbi+1+ sb1+2 for in-2 to 0

The quadratic factor is introduced to allow the determination of complex roots. If the
coefficients are real, of complex roots. If the coefficient are real, the complex roots occur in
conjugate pair. If x2 - rx - s is an exact divisor of the polynomial, complex roots can be determined
by the quadratic formula. This the method reduces to the determining the values of r and s that
make the quadratic factor an exact divisor. In other words, we seek the values that make the
remainder term equal to zero.

In order to make the remainder to be zero, b, and b, must be zero. This method
systematically sugests to modify the initial guesses so that b, and b, approach zero. Note that b,
and b, are functions of both r and s. Using Taylor series expansion, we write

neglecting the second and higher order terms.

Setting the right and side of (2) to zero, we get

From these two equations Ar and As can be calculated. The partial derivatives can be
obtained by a synthetic division of the b's in a fashion similar to the way in which b's themselves
were desired.

c1 = b n

cn-1 = bn+1 + rcn

c1 = b1 + rci+1+ Sc1+2, for i = n - 2, ....................2.1

30

Where

These partial derivatives are obtained by synthetic division of the b's. Substituting in (3),
we get

These equations can be solved for r and s to improve initial guesses of r and s.

when these error estimates are smaller than the prespecified stopping criterion E,, the
values of the roots can be determined by

At this point, there are three posibilities

i) The quotient is a third-order polynomial or greater: Using the previous values of r and s
as starting guesses, new values of r and s would be evaluated.

ii) The quotient is a quadratic: For this case, the remaining two roots can be evaluated
directly with equations (6).

iii) The quotient isa first order polynomial In this case, the remaining single root can be
S
evaluated as x = -
r

Problem 14

Using Bairstow's method, determine the roots of the polynomia

f5(x) = x5 - 3.5x4 + 2.75x3 + 2.125x2 - 3.875x+1.25

with initial guesses r = s = -1 and error es = 1%

 31

Solution

Here a5 = 1, a4 = -3.5, a3 = 2.75, a4 = 2.125, a5 = -3.875, a6 = 1.25

b5 = a5 = 1, r = s = -1

b4 = a4 + rb5 = -3.5+ (-1) (1) = -4.5

Similarly

b3 = 6.25, b2 = 0.375, b1 = -1.05, and b0 = 11.375

Further

c5 = b 5 = 1

c4 = b4 + rc5 = -4.5+ (-1) (1) = -5.5

c3 = 10.75, c2 = -4.875, c1 = 16.375

We know that

Using these equations, we form the following equations

-4.875r + 10.75s = 10.5

-16.375r -4.875s=-11.375

Solving these equations, we get

r=0.3558 and s=1.1381

Thus original guesses can be corrected as

r = - 1 + 0.3558 = -0.6442

s = -1 +1.1381 = 0.1381

Now the approximate errors are

32

Using the revised values of r and s in (*), repeat the above calculations. Then we get b5
= 1, b4 = -4.1442, b3 = 5.5578 b2 = -2.0276, b1 = -1.8013, b0 = -2.1304 c0 = 1, c4 = -4.7884,
c3 = 8.7806, c2 = -8.3454, c1 = 4.7874.

Now we have to solve the following equations

-8.3454 r+8.7806 s = 1.8013

4.7874r-8.3454s = -2.1304

Thus we get r = 0.1331, s = 0.3316. These estimates are used to correct the root
estimate as

r = -0.6442 +0.1331 = -0.5111. = 26.0%

s = -0.1381 +0.3316=0.4697, 70.6%

After four iterations the method converges on value of r =0.5 for which = 0.063%
and s = 0.5 for which

= 0.040%. Now the roots can be obtained using the formula

Now the quotient polynomial can be calculated using synthetic division. At this point, the
quotient is the cubic equation.

q(x) = x3- 4x2 + 5.25x-2.5

Bairstow's method can be applied to this polynomial a(r) using the previous step r = -0.5
and s = 0.5 as starting guesses. After five iterations, we get the estimates of r = 2 and S=-1.249
which can be used to compute
 33

Now we have calculated four roots of the given polynomial. At this point, the quotient is a
first order polynomial that can be directly evaluated by the formula, x = - s/r to determine the fifth
root 2.

Thus we obtain the roots of f(x) as 0.5, -1.01 ± 0.499i and 2.

Check Your Progress

Say True or False

a. Bisection method is simple but slowly convergent method T/F

b. I st order derviate in used in the falsi method T/F

c. this formla used in the Newton's iteration method T/F

1.11 Summary
In this lesson we have learnt to find the roots of equations by using the different methods.

1.12 Check Your Answers

a. True b. False c. True

1.13 Model Questions

1. Determine a real root of f(x) = -2+7x-5x2+6x3 using graphically

2. Using bisection method find a real root of the equations

x3 - 9x - 1 = 0

x3 + 7x - 17 = 0

3. Find a root of the following equations correct to four decimal places using the false
position method.

i) ex sin x = 1 is (0,1) (ans: 5885)

ii) xlog10x 1.2 is (2,3) (ans:2.7406)

iii) x3- 4x - 1 = 0 is (0.1) (ans: 0.2541)

34

4 Using Newton-Raphson method find a root correct to three decimal places of the following
equation:

i) x3-3x-5=0, x0 = 2.0 (Ans:2.279)

i) xe2 - cos x = 0, x0 = 0.0 (ans:0.518)

iii) x sin x + cos x = 0 near x - II (Ans: 2.798)

5. Find a real root of the following equations correct to 3 decimal places by fixed point
iteration method

i) cos x = 3x-1 (Ans: 0.607)

ii) x3 + x2-100 = 0 (Ans:4.331)

iii) sin x = ex - 3x (Ans:0.360)

iv) 1-+ ......= 0 (Ans:1.445)

6. Using the secant method, find a root of the following equations.

i) x2 - 2x - 1= 0, x0 = 2.6, x1 = 2.5

ii) x3 - x -3 = 0, x0 = 1.7, x1- 1.62

iii) x3 - x + 2 = 0, x0 = -1.5, x1 = -1.52

7. Distinguish beween secant and false-position methods.

8. Describe Muller's method for finding a root of a polynomial

9. Explain Bairstow's method for finding a quadratic factor of a polynomial

10. Use Muller's method to determine the real and complex root (if any) of

i) x3- x2 + 2x-2

ii) 2x4 + 6x2 + 8

iii) x4 - 2x3 + 6x2 - 2x + 5

11. Apply Bairstow's method to x4 - 3x3 + 20x2 + 44x + 54 = 0 to find a quadratic factor
close to x2 + 2x + 2 = 0. Do one step only.

(Ans: x2 + 1.9420036x + 1.9543058)

12. Apply Bairstow's method to the x4 + 5x3 + 3x2 - 5x - 9 = 0 to find a quadratic factor close
to x2 - 3x + 5. Do one step only

(Ans: x2-1.8549829x+1.7044674)
 35

LESSON - 2
SOLUTION OF SIMULTANEOUS EQUATIONS
Structure

2.1 Introduction

2.2 Learning Objectives

2.3 Gauss Elimination Methodz

2.4 Gauss Jordan Method

2.5 LU Decomposition

2.6 Matrix Inverse

2.7 Gauss Seidel Iteration

2.8 Summary

2.9 Check Your Answers

2.10 Model Questions

2.1 Introduction
In the previous lessons, we have deterimined the value x that satisfied a single equation
f(x) = 0. Now we learn to determine the values x1, x2, .... xn, that simultaneously satisfy a set of
equations

f1 (x1, x2, .....xn) = 0

f2 (x1, x2, .....xn) = 0

fn (x1, x2, .....xn) = 0

Such system can be either linear or non-linear

Let us study linear algebraic equations that are of the general form
36

a11x1+ a12x2+ ... + a1nxn = b1

a22x1+ a22x2+ ... + a2nxn = b2

an1x1+ an2x2+ ... + annxn = bn

where a's are constant coefficient, b's are constant and n is the number of equations. All
other equations are non-linear.

The study of solution of a system of linear equations is carried out under two different
method: (1) Direct method and (2) Iterative Method.

Direct Methods include Gauss Elimination, Gauss-Jordan Method and LU-Decomposition

(or triangularisation).

2.2 Learning Objectives

After reading this lesson, you must be able to understand:

• the gauss elimination method

• the gauss-Jordan elimination method

• LU decomposition you will know how to incorporate pivoting and also to compute matrix
inverse using triangularisation, Gauss seidal Iteration

2.3 Gauss Elimination Method

An arbitrary system of linear equation may have

i) a unique solution or

ii) an infinite number of solutions or

iii) no solution

i) xy=3
3x + 2y = 7 has a unique solution x = 1, y = 2. This represents a pair of intersecting
straight lines. The point of intersection is the solution.
 37

Scanned with OKEN Scanner

36

ii)

-2x+y=6

4x + 6y = - 12

has infinitely many solutions. They represent the same line. Every

point on the line is the solution.

2x - y = 5

iii)

3x -

= 4 has no solution. They represent a pair of parallel lines and so they

have no point of intersection.

pe senil to mal

Now look at a system of n linear equation in n variables.

bins a11112x2 + ...

nn

+ a1x = a+1

anx1 + a2x2 + ...

38

ii) -2x + y = 6
4x + 6y = - 12 has infinitely many solutions. They represent the same line. Every point
on the line is the solution.

iii) -2x + y = 6
4x + 6y = - 12 has no solution. They represent a pair of parallel lines and so they have no
point of intersection.

Now look at a system of a linear equation in a variables.

a11x1 + a12x2 + .... +a1nxn = a1n+1

an1x1 + an2x2 + .... +annxn = an, n+1

We discuss two methods

1) Direct method (Sweep out methods or Gauss elimination)

2) Iterative method

Triangular form of a system of linear equations:

a11x1 + a12x2 + ... ... + a1nxn = a1’n+1

a22x2 + ... ... + a2nxn = a2’n+1

a33x3 + ... + aanxn=33,n+1

Example

3x1 -2x2- 4x3 = 1

5x2 - 8x3 = -1

4x3 = 8

Gaussian Elimination

• Step 1: Divide the first equation through by the coefficient of x, (If this coefficient happens
to be zero then choose another of the equation and place it first).

• Step 2: Eliminate x, from the second equation by subtracting a multiple of the first equation
from the second equation. Eliminate x1, from the third equation by subtracting a multiple
of the first equation from the third equation.
 39

• Step 3: Divide the second equation through by the coefficient of x, (If this coefficient is
zero then interchange the second and third equations. We shall see later how to proceed
if neither of the second and third equations contain a term in x2). (complication 1)

• Step 4: Eliminate x, from the third equation by subtracting a multiple of the 2nd equation.

• Step 5: Divide the third equation through by the coefficient of x, (if it is zero, we will see
later how to solve the system) (complication 2).

Definition : The number a, in the position (r,r) that is used to eliminate x, in rows r+1, r+2, ..., n
is called the rth pivotal element and the rth row is called the Pivotal row.

Problem 1

2x1 - x2 + 3x3 = 1 (1)

4x1 + 2x2 -x2 = -8 (2)

3x1 + x2 + 2x3 = -1 (3)

(1) ÷ 2 => x1 -1/2 x2 + 3/2 x3 = 1/2 un yo nguoins way sno (ylgium to) abivid

4x1 + 2x2+ x3= -8

3x1 + x2 + 2x2 = -1

(2) - 4 × (1) => x1- 1/2 x2 + 3/2 x3 = 1/2

4x2 - 7x3 = -10 (4)

(3) - 3 × (1) => 5/2 ×2 -5/2x3 = - 5/2 (5)

x1 -1/2 x2 + 3/2 x3 = 1/2

(4) ÷ 4 => x2 - 7/4x3 = -10 (6)

(5) - (6) x 5/2 => 15/8x3 = 15/4 (7)

Now x1 - 1/2x2 + 3/2x3 = 1/2

x2 - 7/4 x3 = -10

x3 = 2

=> x3 = 2, x2 = 1, x1 = -2
40

Method: (Elementary Row operations). In the calculations, we use the following row
operations.

• Omit the unknowns and equality sign

• Interchange rows

• Divide (or multiply) one row through by a number

• Subtract (or add) a multiple of one row from (to) another

Problem 3

x1 +x2 -x3 = 4

2x1- -x2 +3x3 = 7

4x1 +x2 +x3 = 15

Writing 1 1 -1 4

them without 2 3 7

variables 4 1 1 15

1 1 -1 4

0 -3 5 -1

0 -3 5 -1

1 1 -1 4

0 1 -5/3 1/3

0 -3 5 -1

1 1 -1 4

0 1 -5/3 1/3

0 0 0 0 Look here
 41

Conclusion: The system has no solution or they have infinitely many solutions.

Partial Pivoting: Under the rule, when we interchange rows because of a zero coefficient,
we choose to interchange with the row which has the coefficient which is numerically the largest
((i.e.) the largest when the negative sign is discarded)

Problem 3

3x1 -3x2 +x3 = 1 (1)

-x2 +x2 +2x3 = 2 (2)

2x1 +x2 -3x3 = 0 (3)

Convert to an array and proceed:

3 -3 1 1

-1 1 2 2

2 1 -3 0

(1) + 3 => 1 -1 1/3 1/3

-1 1 2 2

2 1 -3 0

1 -1 1/3 1/3

1 -1 1/3 1/3

1 1
(1) + (2) => 0 2 / 2 /
3 3 Look here

(3) - 2 x (1) => 0 3 -11/3 -1/2

1 -1 1/3 1/3

0 3 -11/3 -2/3
Partial Pivoting
1 1
0 0 2 / 2 /
3 3
42

1 -1 1/3 1/3

(2) + 3 0 1 -11/9 -2/9

1
(3) + 2 / 0 0 1 1
3

x3 = 1

x2 = -2/9 + 11/9 = 1

x1 = 1/3 + 1 - 1/3 = 1

x1 = 1,x2 = 1, x3 1

There may be different situation under complication 1 (i.e.) we may find that there is no
equation (row) which can choose which has a non zero coefficient in the appropriate place. In
this case, just move on to next variable.

Problem 4

X1 +X2 -X3 = -3 (1)

2x1 +2x2 +X3 = 0 (2)

5x1 +5x2 -3x3 = -8 (3)

Convert to an array and proceed

1 1 -1 -3

2 2 1 0

5 5 -3 -8

(2) - 2 x (1) => 0 0 3 6

(3) - 5 x (1) => 0 0 2 7

1 1 -1 -3

0 0 1 2
 43

0 0 2 7

1 1 -1 -3

(3) - 2x (2) => 0 0 1 2

0 0 0 3 Look here

There are inconsistent equations. So there is no solution in this case.

Rule: Given a set of (any number of) simultaneous equations in p unknowns:

i. There is no solution if after the Gauss - Elimination process, the last non-zero row has a
1 at the right hand end and zeros elsewhere.

eg: 1 -1 2
0 0 1

ii. There is a unique solution if after the Gauss-Elimination process there are exacting
p non-zero rows, the last of which has 1 in the position second from the right hand end.

eg: 1 2 1
0 1 3

iii) There are infinitely many solutions if after the Gauss - Elimination process. There are
fewer than p non-zero rows and (i) above does not apply.

eg: 1 2 1
0 1 3

Pivoting: The elimination of the n unknown is performed in (n-1) steps. In this process,
at any steps if the pivotal element vanishes or becomes very small compared to the other
element in the pivotal row then the row is interchanged with any other lower row to obtain a non-
vanishing pivotal element or to avoid multiplication by a large number. This strategy is called
pivoting.

Two Types of Pivoting

(1) Partial Pivoting: Let the given equation be (1), (2) ... (n). In the first stage of elimination,
we search for the largest coefficient (in magnitude) occuring in the first column. Suppose it
44

occurs in the k-th equation. This coefficient a,, is then brought as the first pivot by interchanging
equations (k) (1).

In the second step, the second column is searched and the largest (in magnitude) of the
coefficient in all equations except the first is found. The element is brought as the second pivot
by an interchange of the second equation with that having this pivot.

This process is continued untill the final Step.

Complete Pivoting

All the coefficient occuring in the n linear equations are searched for the largest coefficient
(in magnitude) is chosen as the first pivot. The procedure requires not only an interchange of
equation but also an interchange of the position of variable. After this step, elimination procedure
is carried out as under.

Problem 5

Solve the simultaneous equations.

2x1 -2x2 +x3 -3x4 = 2 (1)

x1 -x2 +3x3 -x4 = -2 (2)

-x1 -2x2 +x3 +2x4 = -6 (3)

3x1, +x2 -x3 -2x4 = 7 (4)

Solution: Convert the given system into an array and proceed with array of numbers
only.

2 -2 1 -3 2

1 -1 3 -1 -1

-2 1 2 2 -6

-1 -2 1 2 -6

3 1 -1 -2 7
 45

3
(1) + 2 => 1 -1 1/2 - /2 1

1 -1 3 -1 -2

-2 -2 1 2 -6

3 1 -1 -2 7

1 -1 1/2 -3/2 1

(2) - (1) => 0 0 5/2 1/2 -3

(3) + (1) => 0 -3 3/2 1/2 -5

5
(4) - 3x1 => 0 4 - /2 5/2 4
3
1 -1 1/2 - /2 1

0 -3 3/2 1/2 -5

Interchanging

Row 2 and Row 3

0 0 5/2 1/2 -3
5
0 -4 - /2 5/2 4
3
1 -1 1/2 - /2 1

1 1
(2) + -3 => 0 1 - /2 - /6 5/3

0 0 5/2 1/2 -3

5
0 4 - /2 5/2 4

3
1 -1 1/2 - /2 1

1 1
0 1 - /2 - /6 5/3

0 0 5/2 1/2 -3

1 8
0 0 - /2 10/6 - /3
46

3
1 -1 1/2 - /2 1

1 1
0 1 - /2 - /6 5/3

6
0 0 1 1/5 - /5

49
0 0 0 49/15 - /15

3
1 -1 1/2 - /2 1

1 1 5
0 1 - /2 - /6 - /3

6
0 0 1 1/3 - /5

0 0 0 0 -1

x1 = 1, x2 = 1, x3 = -1, x4 = -1.

Conditioned System

A system is well-conditioned system if small change in one or more of the coefficient will
result in small change in the solution.

Ill-conditioned systems are those where small changes in coefficient result in large change
in the solution.

Problem 6

Solve the following system

+2x2 = 10 (1)

1.1 x1 + 2x2 = 10.4 (2)

The solve it again but which the coefficient of x1 in (2) is modified as 1.05

Solution

Solving (1) and (2), we get

2(10) - 2(10.4)
x1 = =4
1(2) - 2(1.1)

-
 47

1(10.4) 1.1(10)
x2 = =3
1(2) - 2(1.1)

However with the change in (2) from 1.1 to 1.05, we get

2(10) - 2(10.4)
x1 = =8
1(2) - 2(1.1)

1(10.4) 1.05(10)
x2 = =1
1(2) - 2(1.05)

There is drastic change in the result.

2.4 Gauss Jordan Elimination Method

The coefficient matrix is reduced to a diagonal matrix rather than a triangular matrix. At
all steps of the Gauss Elimination method, the elimination is done not only in the equation solved
but also in the equation above, producing the solution without using back substitution method.
On the completion of Gauss Jordan method, the equation AX = B becomes

The Solution is given by x1 = d1, i = 1,2,..., n

i.e. [A|b] --->GJ ? [I|d]

Problem 7

Find the inverse of the coefficient matrix of the system

1 1 1 x1 1

4 3 -1 x2 = 6

3 5 3 x3 6

by Gauss Jordan method with partial pivoting and hence solve the system.

Solution

Use the augmented matrix [A|l], We get

48
 49

2.5 LU Decomposition Method

Let AX = b. The coefficient matrix A is decomposed or factorised into the product of a
lower triangular matrix L and an upper triangular matrix U.

A=LU

where
50

where u1 = 1.

Note: A-1 can be determined as A-1 = U-1 L-1

Problem 8

Use the decomposition method to solve the system

 51

Therefore,

1 1 1 1 0 0 1 1 1

4 3 -1 4 -1 0 0 1 5

3 5 3 3 2 -10 0 0 1

A=LU

becomes AX = b LUX = b, ==> UX = Z and LZ = b.

For UX = Z we get
52

1 1 1 x1 z1

0 1 5 x2 = z2

0 0 1 x3 z3

and

1 0 0 z1 1

4 -1 0 z2 = 6

3 2 -10 z3 4

From LZ = b, using formal substitution, we get z, = 1 4z1 - z2 = 6 ==> 4 - z2 = 6 (i.e.) z2

= -2
 53

Problem 9

Employ LU decomposition to determine the matrix inverse for the matrix.

Solution

The LU decomposition of A can be calculated to be

and

The first column of A-1 is calculated as follows. Perform forward substitution solutio
procedure with a unit vector (with 1 is the first row) as the right-hand side vector

Thus we get DT = [1 -0.033333 -0.1009]. This vector can be used as the right han side
of UX = D (i.e)

which can be solved by back substitution for XT = [0.33249 -0.00578-0.01008] which is

the first column of the matrix.
54

To determine the second column of A1, solve the following

As before, from this we get XT = [0.004944 0.142903 0.00271] which is second column
-1
of A (i.e.)

Finally, the forward and back substitution procedures can be implemented with BT =
[0,0,1] to solve for

XT = [0.006798 0.004283 0.09988],

which is the final column of the matrix.

Error Analysis: The inverse provides a means to discern whether systems are ill-
conditioned. Three methods are available for this purpose.

1. Scale the matrix of coeffcients A so that the largest element in each row is 1. Invert the
scaled matrix. If these are elements of A-1 that are several orders of magnitude greater
than one, it is likely that the system is ill-conditioned.

2. Multiply the inverse by the original coefficient matrix and assess whether the result is
close to the identify matrix. If not, it indicates ill-conditioning.

3. Invert the inverted matrix and assess whether the result is sufficiently close to the original
coefficient matrix. If not, it again indicates that the system is ill-conditioned.
 55

2.6 Matrix Inverse

Doolittle Decomposition

In the LU decomposition, the L matrix has 1's on the diagonal.

Crout decomposition

In the LU decomposition, the U matrix has 1's on the diagonal.

Matrix Inverse Through Crout Decomposition

We have A = LU where 1's on the diagonal of U. Then

A1 = (LU)-1 = U-1 L-1

Since L and U are known, their inverses L-1 and U-1 can be determined using LL-1 = | and
UU-1 and I and hence A-1 = U-1 L-1 can be computed.

Problem 10

Find the inverse of by Crout's method.

Solution

Let A =

The matrix A can be decomposed into LU as

A = LU =
56

Hence |11 = 2, |21, = 2, |31 = -1

U12 = -1, U13 = 2, |22 = 5, |32 = 0

2
U23 = - /5 and I33 = 1 (using Crout's method).

Hence
 57

This Gives

2.7 Gauss-Seidel Iteration Method

Consider the following system of equations.

a11x1 + a12x2 + a13x3 = b1

a21×1 + a22x2 + a23x3 = b2

a31x1 + a32x2 + a33x3 = b3

This method is applicable only when each equation of the system contains one coefficient
much larger than the others in that equation and the larger coefficient in different equations
correspond to the different variables (unknowns). Arrange the system in such a manner that the
largest coefficient lie on the diagonal.

Assume that a11, a22, a33 are the largest in its rows of the given system. Write the system
as

First Iteration

Start x2 = 0 and x3 = 0. Then

1
x1(1) = /a11 b1
58

Second Iteration

Third Iteration

and so on untill we get the roots within the desired degree of accuracy.

Problem 11
 59

Iteration x1 x2

1 1 2.25

2 0.4375 1.9688

3 0.5078 2.0039

4 0.4990 1.9995

5 0.5001 2.000

6 0.5000 2.000

|x1(4) - x1(5)| < 10-4 Therefore x1 = 0.5 x2 = 2.0000

Problem 12

Show that the Gauss-Seidel method for solving

a11x1+ a12 x2 = a13 (1)

a21x1 + a22x2 = a23 (2)

will converge if a12a21| > |a11a22|

Proof: Rewrite the given equations as

x1 = (a13 - a12x2) / a11 (3)

x2 = (a23 - a21x1) / a22 (4)

x1(0) = x1 = 0x2(0) = x2 = 0. Putting x2(0) = 0 in (3), we get

First Iteration

x1(1) = (a13- a12.0) a11

x2(1) = (a23a21X,(1) a22

Second Iteration

x1(2) = (a13 - a12 x2(1) / a11

x2(2) = (a23 - a21 x1(1) / a22

60

(k+1)st Iteration

x1(k+1) = (a12 - a12x2(k) / a11

x2(k+1) = (a23 - a21x1(k+1) / a22

If x2 (k+1) = x2(k) (or approx. equal), the procedure is stopped and the values of x, and are
obtained as x1 = x1(k+1) and x2y = x2(k). Then the iteration procedure converges.

If x2(k) x2(k), we have to continue the iteration.

Is it good to continue?

Conditions for convergence

At (k+1) iteration: Recall

X(k+1)1 = (a13 - a12x(k)2) / a11 (1)

x(k+1)2 = (a23 - a21x1(k+1)) / a22 (2)

At the (k+2) iteration, we get

x1(k+2) = (13 - a12x2k+1) / a11 (3)

x2(k+2) = (a23 - a21x1(k+2)) / a22 (4)

Eliminating x,(k+1) from equations (1) and (2), we get

a11
/ a (-a22x2(k+1) + a23) = a13 - a12x2(k) (5)
21

Similarly eliminating x1(k+2) from (3) and (4), we get

a11
/ a (-a22x2(h+2) + a23) = a13 - a12x2(k) (6)
21

(5) and (6) ==>

a21
x2(k+2) - x2(k+1) = / a a x a12 (x2(k+1) - x2(k))
21 22

a12a21
d2k+1 = / d2(k)
a11a22
For the convergence in the above procedure, we must have

d2(k+1) < d2(k)

 61

This will ensure that x, tends stabilize at certain steps and this in turn gives a stable
value x,.

Thus the condition for convergence is

i.e. the diagonal element is larger than the remaining elements.

General case (n equations in n unknowns)

and

for atleast one i (i.e diagonally dominant coefficients)

Consider the system of equations.

Problem 13

Interchanging (1) and (3), we get

Now Gauss-Seidel is applicable!

62

Example

Consider the system of equations.

Here it is not possible to effect such an interchange in every case. So Gauss-Seidel

procedure does not lead to a favourable solution set.

Problem 14

Solve, by Gauss-Seidel method, the following system of equations:

Solution

The coefficient matrix of the given system is not diagonally dominate. So, rearrange the
equations so that the elements in the coefficient matrix are diagonally dominant.
 63

Second iteration leads to

After fifth and sixth iterations, we get

x5 = 0.9935893

y5 = 1.5069741

z5 = 1.8485488

and x = 0.9935947

y6 = 1.5069774

z6 =1.8485473

respectively.

Since the values of x,y and z in the fifth and sixth iterations are same up to four decimals,
we stop the iteration process.

Thus we get x = 0.9935y = 1.5069z = 1.8485

64

Check Your Progress

Fill in the blanks

1. Gauss-Seidel________________method

2. Gauss Jordon _______________methoa

2.8 Summary
In this lesson we have learnt about Gauss elimination, Jordon and Gauss-Seidal Iteration
methods.

2.9 Check Your Answers

1. Iteration

2. Elimination

2.10 Model Questions

1. Explain Gauss elimination method

2. Describe Gauss-Jordon elimination method.

3. Explain LU decomposition in matrix analysis.

4. Explain (i) Partial pivoting (ii) complete pivoting

5. Explain how matrix inverse helps to analyse errors in the system.

6. Use Gauss elimination to solve the following system of equations.

(i) 3x + y + z=3

2x - 8y + z = -5

x-2y+9z = 8

(Ans: x = y = z = 1)

ii) x1 + x2 + x3- x4 = 2

x1 + x2 + 3x3 + 2x4 = -6
 65

2x1 + 3x2 - x3 + 2x4 = 7

x1 + 2x2 + x3 - z4 = -2

7. Use Gauss-Jordan methods to solve the following system of equations.

i) 10x + y + z = 12

2x + 10y + z = 13

x + y + 5z = 7

(Ans: x = y = z = 1)

ii) x + 2y + z = 8

2x + 3y+4z = 20

4x + 3y + 2z = 16

(Ans: x = 1, y= 2, z=3)

8. Solve the following system of equations by the method triangularisation (or LU

decomposition)

i) 2x-3y+ 10z = 3

-x + 4y + 2z = 20

5x + 2y + z = -12 =

(Ans: x = -4, y = 3, z= 2)

ii) 3x + 2y + 7z = 32

2x + 3y + z = 40

3x+4y-z = 56

(Ans: x = 7, y = 9, z = -1)
66

9. Find the inverse of the following matrices using triangularisation method A =

10. Find the inverse of the matrix A using Crout's Method A =

 67

LESSON - 3
NUMERICAL INTEGRATION
Structure

3.1 Introduction

3.2 Learning Objectives

3.3 Trapezoidal Rule

3.4 Simpson's Rule

3.5 Romberg Integration

3.6 Summary

3.7 Check Your Answers

3.8 Model Questions

3.1 Introduction
Differentiation and integration are two important mathematical concepts. They are related
to find a tangent to a curve and area under the curve between two points respectively.
Mathematically inverse process of differentiation is the integration.

The function to be differentiated or integrated will typically be in one of the following three
forms.

i) A simple continuous function such as a polynomial, an exponential, or a trigonometric

function

ii) A complicated continuous function that is diffcult or impossible to differentiate or integrate

analytically.

iii) A tabulated function where values of x and f(x) are given at a number of discrete points
as is often the case with experimental or field data.

We learn about some Newton-Cotes formulae for numerical integration viz. Trapezoidal
rule and Simpson's Rule.
68

The main idea of numerical integration is based on replacing a complicated function or

tabulated data with a simple polynomical that is easy to integrate.

Quadrature. The process of computing f(x) dx using a set of tabulated values (x,
f(x1)),i= 0, 1, 2,....n, a = x0 and b = x0 is called numerical integration. Since f(x) is a single variable
function, the process in general is known as quadrature.

Closed forms of Newton-Cotes formulae

The closed forms are those where the data points at the beginning and end of the limits
of integration are known.

Open forms of Newton-Cotes formulate

The open forms have integration limits that extend beyond the range of data. They are
not used for definite integration.

3.2 Learning Objectives

After studying this lesson, you will be able to integrate numerically using

• trapezoidal Rule

• simpson's 1/3 Rule and 3/8 Rule

• romberg intergration

3.3 Trapezoidal Rule

The trapezoidal rule is the first of the Newton-Cotes formulae. In general Newton- Cotes
formula has the following principle.

I= f(x) dx ~ pn(x) dx (1)

where p (x) = a polynomial of the form

a0 + a1 x + ... + anxn,

n being the order of the polynomial.

 69

In the above representation, take n = 1 (straight line). Then

where the integrated refers to a stright line. The area under this straight line is an estimate
of the integral of f(x) between the limits a and b.

Now

which is the formula for the trapezoidal rule.

3.3.1 Composite Trapezoidal Rule

b-a
Divide (a,b) into n equal subintervals, each subinterval of width h = , If a and b are
n
designated as x0 and xn, then

where a = x0 < x1 < x2 < ... < xn = b. Substituting the trapezoidal rule for the each integral,
we get
70

Errors of the Trapezoidal rule are calculated to be

i) Simple case: E1 =

ii) Composite case:

3.4 Simpson's - Rule

Now take n = 2 in (1) is the interpolating polynomial is of second order.

If a and b are designated as x and x, and f(x) is represented by a second order Lagrange
polynomial, the integration becomes

After integration and simplification, we get

1~ h/3 [f(x0) + 4f(x1) + f(x2)]

 71

where h = b - a / 2 This is known as Simpson's 1/3 Rule. It is the second Newton-Cotes

closed integration formula.
(b-a)5
Error is calculated to be E, = = f(4) (x)
2880

3.4.1 Composite Simpson's 1/3 - Rule

6-a
Divide the interval (a,b) into a number of segments of equal width h = total integral
n
can be represented as

Substituting Simpson's 1/3 - Rule for the individual integral gives

Thus we get

Notice that an even number of segments must be utilized to implement the method.

Error estimate in composite Simpson's 1/3 Rule in calculate to be E~ where

(4) is the average fourth derivative for the interval.

3.4.2 Simpson's 3/8 - Rule

In a similar manner to the derivation of the trapezoidal and Simpson's 1/3 third order
Lagrange polynomial can be fit to four points and integrated.

gives
72

where h= b-a This is the third Newton-Cotes integration formula. The 3/8 be expressed in the
form.

with errors.

This rule is preferred if the number of segments is odd. Usually Simpson's 1/3 Rule is
preferred because it attains third order accuracy with three points rather than four points required
for 3/8 rule.

3.4.3 Choice of the Rule

For even number of segments, Simpson's 1/3 Simpson's rule is preferred. Suppose
that we desired to estimate for five segments. One choice would be to use a composite trapezoidal
rule. This is not advisable, however, because of the large truncation error associated with this
method. An alternative would be to apply Simpson's 1/3- Rule to the first two segments and
Simpson's 1/3 Rule to the last three. In this way we could obtain an estimate with third order
accuracy across the entire interval.

Problem 1

Use simple Trapezoidal rule to integrate f(x) = 0.2 + 25x-200x2 + 675x3 - 900x4 + 400x5
from a = 0 to b = 0.8.

Solution

We have f(0) = 0.2, f(0.8) 0.232

Substituting these values in I = (b- a) we get

 73

Problem 2

Evaluate by using

i) Trapezoidal Rule

ii) Simpson’s 1/3 Rule

iii) Simpson’s 3/8 Rule

Solution

Take h = 1. Divide the whole range [0, 10] into 10 equal points. The values of the integral
for each point of sub-division are given below.

X y = f(x) x y = f(x)

0 1 6 0.27027

1 0.5 7 0.02

2 0.2 8 0.0153846

3 0.1 9 0.0121951

4 0.0588235 10 9.9009901 x 10-3

5 0.0384615

i) Trapezoidal Rule

I =

= h \ 2 [(y0 + y10) + 2(y1 + ... y0)]

= 1/2 [(1+9.9009901 x 10-3)] + 2(0.5 +0.2 + ... + 0.0121951)]

= 1.4768422.

ii Simpsons 1/3 Rule

= 1.4768422.
74

= 1.4316659 (after Calculation)

ii) Simpson's 3/8 Rule

= 1.4198828 (after calculations)

3.5 Romberg Integration

Romberg integration is one technique that is designed to attain efficient numberical
integrals of functions. It is based on the successive application of the Trapezoidal Rule.

For an interval of size h, the error in the trapezoidal rule ismul eat gigge
 75

We have obtained this result by applying Trapezoidal rule twice. By applying the rule
several times, every times having h, we get a sequence of results L1, L2,..... in which the error
is reduced by 1/4 every time.

We apply the formula (*) again to each pair of L,'s ie L1, L2, L3, L4; ..... to get improved
results M1, M2, M3.....

Pick pairs from M's and apply (*) to get improved results N1, N2,

Repeat this calculation until two successive values are close to each other.

This method is called Richardson's extrapolation and its systematic improvements is

called Romberg Integration.
76

Problem 3

Now extrapolate

Improved I =

= 0.65947 + 1/3 (0.65947 -0.66211)

= 0.65859

We tabulate the values of Romberg integration as follows

 77

Improved estimates

0.66211

0.65947 0.65859

0.65898 0.65881 0.65882

0.65886 0.65882 0.65882 0.65882

Finally dx~0.65882.

Romberg Method for a Tabulated Function

We can apply the Romberg method to integrate the function known only as a table of
evenly spaced function values, but now we cannot make h smaller. Instead we use estimates of
the integral with h doubled each time.

Problem 4

Use the data in the following table to get the integral between the limits x = 1.8 and x = 3.4
Begin with h = 0.4.

X f(x) X f(x)

1.8 6.050 2.8 16.445

2.0 7.389 3.0 20.086

2.2 9.025 3.2 24.533

2.4 11.023 3.4 29.964

2.6 13.464

Solution

The formula in this case is First estimate is

78

I1 = 0.4/2 [6.050 + 2(9.025) + 2(13.464) + 2(20.086) + 29.964]

= 24.2328

Use now h = 0.8 to get the next extimate.

I2 = 0.8/2 [6.050+2(13.464) + 29.964]

= 25.1768

Now extrapolate

I3= 24.2328+ 1/3 [24.1768-25.1768]

Had we started with h = 0.2 we would have had the results shown in the following table.

h = 0.2 23.9944 23.9149 23.9147

h = 0.4 24.2328 23.9181

h = 0.8 25.1768

Thus the approximate value of the integral is 23.9147.

Check Your Progress

Fill in the blanks

1. Trapezoidal Rule in..........................and .................... of all the methods.

2. Y = f(x) then I ~ 3h/8 (yo + yn + 32 yn-1 + 2 Eyn-3] this is Simpson's Rule

3.6 Summary
In this lesson we have learnt some important numerical Integration methods.

3.7 Check Your Answers

1. Less accurate, simplest

2. 3/8
 79

3.8 Model Questions

1. Derive Trapezoidal rule, simpson's 1/3, 3/8 rule.

2. Describe the method of Romberg Integration.

3. Find the approximate value of by 1/3, 3/8 rule using 8 equals parts.

4. Find the approximate value of [Link] by (i) Trapezoidal, simpson's 1/3 Rule, (

ii) Simpson's 3/8 Rule using ordinates.

5. Calculate the trapezoidal approximation, and the simpson's rule approximation of the

integral |= with h = 0.5 and 1.0 comment on the result.

6. Use Romberg's method to evaluate correct to three decimal places with h =

0.5, 0.25, 0.125.

80

LESSON - 4
DIFFERENTIAL EQUATIONS
Structure

4.1 Introduction

4.2 Learning Objectives

4.3 Taylor's Method

4.4 Euler's Methodicblasental gtx xne

4.5 Rurge Kuha Method

4.6 Predictor - Corrector Method

4.7 Summary

4.8 Check Your Answers

8.0 4.9 Model Questions

4.1 Introduction
Differential equations involving only one independent variable are known as ordinary
differential equations.

dy

-X

dx

dy

dx

are example of ordinary differential equations.

 81

If the dependent variables are functions of more than one independent variables, the
equations are known as parital differential equations. For example,

is a partial differential equation. In this lesson we shall consider only ordinary differential
equations and their numerical solutions.

An ordinary differential equation may involve one or more derivatives. If an equation

contains nth derivative, then it is said to be of order n. For example

+ y = 0 is a second-order differential equation.

First-order differential equations can be written in the form

y = f(x,y)

Higher-order equations can be reduced to a set of first-order equations by suitable

transformations. For example, consider a second-order equation of the type

y" = f(x, y, y)

Let us suppose

y' = u

Then y" = u'

and the equation becomes

u' = f(x, y, u)

y' = u

4.2 Learning Objectives

After studying this lesson, you should be able to understand the following methods.

• Taylor's method

• Euler's method

• Runge - Kutta method

• Prdictor - Corrector method

82

4.3 Solution by Taylor Series

We can expand a function y(x) about a point x = x0 using Taylor's theorem of expansion
yn(x0) yn(xo)
y(x) = y(xo) + (x - x0) y(x0) + (x - x0)2 / + ...... + (x - x0)n / where y' (x0) is the
21 n!
ith derivative of y(x) evaluated at x = x0.

The value of y(x) can be obtained if we know the values of its derivatives. This implies
that if we are given the equation

y' = f(x, y)

we must then repeatedly differentiate f(x, y) implicitly with respect to x and evaluate them
at x0.

For example, if y' = f(x, y), then

where denotes the function f(x, y) and f, and f, denote the partial derivatives of the function
f(x,y) with respect to x and y respectively. Similarly, we can obtain

yn = fxx + [Link]+ [Link] + f. f2y

Problem 1

Consider the equation

y1 = x 2 + y 2

unter the condition y(x) = 1 when x = 0

y1 = x 2 + y 2

yn = 2x + 2yyn
 83

yn = 2 + 2yyn +2(y)2

At x = 0, y(0) = 1 and therefore

y' (0) = 1

y" (0) = 2

y" =2+2yy" + 2(y')2

At x = 0, y(0) = 1 and therefore

y' = (0) = 1

y"(0) = 2

y'''(0) = 2 + (2)(1)(2) + (2)(1)2 = 8

Substituting these values, the Taylor series becomes

y(x) = 1 + x + x2 + 8/31 x3 + . . .

The number of terms to be used depends on the accuracy of the solution needed.

Problem 2

Using Taylor's series method compute y(0.1) correct to 4 decimal places. If y(x) statisfies
1
y = x + y, y(0) =1.

Given y1 = x + y and x0 = 0, y0 = 1 we know that the Taylor series formula for

y"(x0)
y(x) = y(x0) + (x - x0) y1(x0) + (x - xo)2 / + .... + (x-xn)n yn(xn)
2!
Given condition y(x) = 1 when x0 = 0

y' = x + y : x 0 + y0 = 0 + 1 = 2

y" = 1+ y': 1+ y01 = 1+1=2

y"" = y": y011 = 2.

2 (x-0)3 2
y(x) = 1 + (x -0). 1 + (x -0)2 / + /
2! 3!
84

= 1 + x + x2 + x3 1/3 + ..................
(0.1)2
y(0.1) = 1 + 0.1+ (0.1)2 + / 2 +.....
3!
= 1 +0.1+0.01 +0.0003

= 1.1103

~y(0.1) = 1.1103

4.4 Euler's Method

Euler's method is a step-by-step iterative method with the initial conditions as the starting
point. Consider the equation

y' (x) = f(x, y)

The mean value theorem states that if a function is continuous and differentiable between
two points on the function, then the slope of the line joining these two points is equal to the
derivative of the function at least one other point between these two points (Fig. 4.1). That is for
x1 < a < x2, we have
y(x2) - y(x1)
y'(a) =
x2 - x1

Fig. 4.1 Illustration of Mean Value Theorem

 85

Letting a = x1 and x2-x1 = h,

We have

y(x2) - y(x1) = hy'(x1)

Substituting for y'(x1) and rearranging, we get

y(x2) = y(x1) + hf (x1, y1)

Similarly we get

y(x3) = y(x2) + hf/2 (x2 y2)

In general,

yi+1= y1 + hf/2 (x1, y1) i = 1, 2, 3...

This is known as Euler's formula can be used repeatedly to obtain y(x) for any value of x.
In order to improve accuracy, the value of h should be as small as possible.

Suppose the initial value is x, and we want the function at x = xn. The have to divide the
range x, to xn into (n-1) equal intervals of width h. Then
xn - x1
h=
n-1

4.4.1 Improved Euler's Method

Le the given first order differential equation be

dy
/ = f(x, y) (1)
dx

Let us the solve this equation under the condition y(x0) = yo

Starting with initial value yo, an approximate value for y, is compute from the relation

y1(1) = y0 + f(x0, y0) h (2)

dy
Substituting this approximate value of y, in (1) we get an approximate value of / at
dx
(x1 y1)
86

i.e., dy/dx = f[x1, y1(1)]

Now an improved vlaue of y, is found by multiplying h with the mean values of dy/dx at x0
and x1.

In general

yi+1= y1 + h f(x1 + h/2, y1+h/2. f(x1, y1)}

This formula is called improved Eulers' Formula

4.4.2 Modified Euler's Method

In general

yi+1 = y1, + h f[x1 + h/2, y1 + h/2 f(x1, y1)]

(or)

(x+h) = y(x) + h f[x + h/2, y + h/2 f(x, y)

Problem 2

Using Improved Euler's Method

dy 2x
Find y at x = 0.1 and x = 0.2 given /dx = y - /y, y(0) =1
 87

Solution

Substituting (3) in (2) we get

Sustituting (5) in (4) we get

88

Substituting (7) in (6) we get

Substituting (9) in (8) we get

y2 = 1.0959 +0.05 (0.9135 +0.8503)

y(0.2) = 1.1841

X 0 0.1 0.2

y 1 0.959 1.1841

4.5 Runge-Kutta Method

Runge-Kutta method is more accurate as compared to the methods discussed so far.
In this method, the integration interval is divide into subintervals and a weighted average of
derivatives at these intervals is used to determine the value of the dependent variables. One of
the advantages of this method is that it is a out-step method. That is, in order to evaluate y,,, we
need information only at the preceeding point x1, y1.
 89

Consider Euler's modified formula (predictor-corrector formula) again.

y2 = y1 + h/2 [f(x1, y1) + f(x2, y2)]

From Euler's formula, we know that

y2 = y1 + hf(x1, y1) = y1+ hf1

If we substitute this for y, on the right-side of Euler's modified formula, we get

y2 = y1 + 1/2 [h.f 1 + hf(x1 + h, y1 + hf1)]

where f1 = f(x1, y1)

Letting k1 = hf1

k2 = hf(x1+ h, y1 + hf,)

We get

y2 = y1 + 1/2 (k1+ k2)

This equation is known as second-order Runge-Kutta formula. It is also known as (OS)

P Heun's method.

To reduce the errors further, we have third-order and fourth-order Runge-Kutta methods.
We shall now consider the fourth-order method.

The fourth-order Runge-Kutta formula contains four parameters and it defined as

y2 = y1 + 1/6 (k1 + 2k2 + 2k3 +k4)

and the parameters k1's are given by

k1 = hf(x1, y1)

k2 = hf(x1+h/2, y1 + k1/2)

k3 = hf(x1+ h/2, y, + k2/2)

k4 = hf(x1+ h, y1 + k3)
90

Problem 3

By applying the fourth order Range Kutta Method find y(0,2) from

y1 = y - x, y(0) = 2 taking h = 0.1

Solution

Given y1 = y -x ie f(x, y) = y - x and y(0) = 2 ie x0 = 0, y0 = 2 and h=0.1

We known that the fourth order Runge - Kutta formula for finding the Ist increment in
y viz Ay is given by
91
92

Hence we have the following table.

X 0 0.1 0.2

y 2 2.20517 2.42139

4.6 Predictor - Corrector Method

In Euler's method we used y'(x,) to approximate y'(a). Another possible approximation is
to used the mean value of the derivatives at x1 and x2 to replace y'(a). That is
 93

Then, we have

Therefore

Notice that this equation contains the term yi+1 on both the sides and, therefore, it is not
possible to evaluate it in the present form. We rewrite this equation as follows:

yi+1c = y1 + h/2 [f(x1, y1) + f(xi+1, yi+1,p)]

This equation is known as corrector formula and y,,,,c is called corrected value y11, P is
called the predicted value and is predicted by Euler's formula

yi+1 p = y1 + hf(x1, yj)

This equation is known as the predictor formula.

The predictor - corrector method is, thus, a two-step method.

Step 1: Predict y, using the predictor formula.

Step 2: Correct y, using the corrector formula.

Step 2 wil be executed repeatedly until the difference two successive value of Y1.1 agree
within some predefined limit.

The entire process will continue till x reaches the desired value.
94

Check Your Progress

Choose the Right Answer

1. The accuracy of the solution is depends upon the number of terms.

a) Talyor b) Euler c) Runge Kutta

2. The value of h should be small the corresponding the value of accuracy should be improve.

a) Talyor b) Euler c) Runge - Kutta

4.7 Summary
In this lesson we have learnt some important numerical solution of differential equation
methods.

4.8 Check Your Answers

1. a 2.b

4.9 Model Questions

1. Solve the following differential equations using Euler's method for 1>, x > 0 Tabulated the
results for x = 0, 0.2, ... 1.0 for various integrations intervals.

a) y' = 2xy y(0) = 1

b) y' = x-y y(0) = 1

2. Solve the differential equation f(x,y) = x + y by Runge - Kutta 4th order method at x=0.1,
0.2, 0.3 given x = 0 when y = 1, y = 0.1

3. Solve the initial value problem dy/dx =3x-4y, y(0)=2 at x=0.4 taking h=0.2 by Runga-
Kutta method of fourth order.

4. Using Taylor series for y(x) find y(0.1) correct to four decimal places if y(x) satisfy the
equation y1 = x - y2 and y(0) = 1.
 95

LESSON - 5

DIAGRAMMATIC AND GRAPHICAL

REPRESENTATION
Structure

5.1 Introduction

5.2 Learning Objectives

5.3 Types of Diagrams

5.4 Frequency Distribution

5.5 Graphs of Frequency Distribution

5.6 Summary

5.7 Check Your Answers

5.8 Model Questions

5.1 Introduction
We know that Statistics is the science which deals with numerical data, (i.e., information
which are expressed in quantities) their analysis and interpretation. Collection of data is the
basis for any statistical inquiry. The huge data collected are arranged in a systematic manner to
enable the statistician to compare, analysis and interpret the data. This process of arranging the
data is known as Classification. Tabultation is the next process. In this the classified data is
presented in a tabular form to enable comparison between the figures. To understand the data
easily we draw diagrams and graphs. Diagrams have the advantage that they are more attractive
and appealing to the eyes. Even lay man can understand the data if presented in the form of a
diagram and it require lesser time to grasp the significance of the data, if presented in the form
of figures.

5.2 Learning Objectives

After studying this lesson you should be able

• to draw Statistical diagrams and graphs

96

• to compare different statistical diagrams

• to form the cumulative distribution. in

5.3 Types of Diagrams

Some of the important statistical diagrams are (1) Bar diagram (2) Pie diagram (3)
Pictogram

5.3.1 Bar Diagram

Bar diagram is the simplet and of all statistical diagrams. It consists of number of rectangle
of equal width and their length are proportional to the magnitude of the variables they represent.
They all stand on the same base and each separated by equal distances. The rectangles may
be either horizontal or vertical.

Problem 1

The following data relate to the enrolment of student to a university. Draw a bar diagram
for the data.

Year No. of Students

1950 3650

1955 5100

1960 9662

1965 10395
 97

5.3.2 Subdivided or Component Bar Diagram

If the variables are subdivided to two or more subvariables, then the corresponding
rectangles are subdivided into parts to represents the subvariables. The lengths of the parts are
proportional to the magnitude of the subvariables. To distinguish the various components we
given different shades. The resulting diagram is known as component bar diagram.

Problem 2

The following data relate to the number of male and female students of three college in
a year. Draw a suitable diagram,

College Number of Students

Men Women

A 750 500

B 800 600

C 1000 900

D 1200 1000

To enable comparison between two variables easily we draw adjacent bars corresponding
to the variables. The resulting diagram is known as super imposition of bar diagram.
98

Problem 3

Draw a component bar diagram for the following data regarding the studnet enrolment of
three Universities.

University Enrolment Total

Arts Science Professional

Courses

A 10000 25000 45000 80000

B 7500 14000 28000 49500

C 12500 18000 36000 66500

Scale: 1cm = 10000 students

 99

5.3.3 Percentage Bar Diagram

If the subvariables are expressed as percentage of the variables then the resulting
component bar diagram is called Percentage bar diagram. Since the total for all the variable are
100, the rectangle of the different variables of equal length and width.

Problem 4

Draw the percentage bar diagram for the data in problem 3.

University Percentage of Students Total

Arts Science Professional

Course

A 12.50 31.25 56.25 100

B 15.15 18.28 56.57 100

C 18.50 27.10 454.10 100

5.3.4 Pie Diagram

Instead of representing the variable by means of rectangles, we can represents them by

means of circles. The circles of drawn such that their areas are proportional to the magnitude of
the variables they represent. Hence the radius of the circles are proportional to the square root
of the magnitude of the variables they represent. To represent the subvariables, the circles are
100

subdivided into sectors such that their areas are proportional to the magnitude of the subvariables.
That is the angles of the sectors are proportional to the magnitude of the component parts.

Problem 5

Draw circular diagram for the following data. of monthly expenditure of two families.

Item Expenditure in Rs.

Family A Family B

Food 300 500

Rent 250 400

Clothes 100 150

Education 120 200

Saving 90 160

Miscellaneous 40 90

Total 900 1500

Draw two circles of radius/900 and 1500 i.e; 30 and 38.7

Calculation of Angles

Family A Family B

Food

Rent

Clothes

Education

Saving

Miscellaneous
 101

5.3.5 Pictogram

A pictogram is a diagram in which picture are drawn. Each full picture represent a certain
stated number. Fraction of the numbers are represented by parts of the figure. Suppose the
enrolment of student of a colleges is as follows:

Year Enrolment
1985 1000
1986 1450
1987 1600
1988 1800

assuming that each picture a man represent 200 students.

1985     
1986     
1987     
1988     
5.3.6 Three Dimensional Diagrams

Three dimensional diagrams are not generally recommended. The cube is the easiest
three dimensional diagram. The sides are proportional to the cube root of the magnitude of the
variables. The choice of the diagram to be drawn for a data depends on the nature of the data
and also the person for whom it is meant. If the person is not literate pictogram and pie diagram
are most suited. For thses who read are understand numbers other types of diagrams are
useful.
102

Check Your Progress - 1

Fill in the blanks

a. A picture is worth ________________________ words.

b. Bar diagrams are ________________________ dimensional diagram

C. A pie diagram is a circle broken down into components section T/F

5.4 Frequency Distribution

A frequency distribution is obtained when the data is grouped into different classes of
appropriate size giving the number of observation in each class. The number of observation of
any class is known as the frequency of the class.

5.4.1 Class Interval

The difference between the highest and lowest value of a class is known as the Class
Interval, For example the class interval of the class 20-30 is 10. W have to select the class
interval having the following two points in mid.

(1) In Calculating statistical constants, we assume that the value of a class is equal to
the mid value of the class interval. If the class interval is too wide the error due to our assumption
will be very large. Where as if the class interval is too short there will be more number of classess
and the calculation will be tedious. So we have to select the class interval in such a way that we
get approximately 10 to 15 classes. H.A. sturges had given a formula for the class intervals as
= Range / 1+ 3.322 log10. . If the calculated value is a fraction, we may take the class interval to
be the integer close to the calculated value. The class intervals need not be uniform. For example
in the data regarding in income tax payees, for different slabs, the class intervals are not uniform.
Wherever possible uniform class interval is preferable.

5.5 Graphs of Frequency Distribution

When a frequency distribution is represented graphically we get a frequency graph.
There are four frequency graphs.

1) The Line Diagram

2) The Rectangle Histogram

3) The Frequency Polygon

4) The Frequency Curve.

 103

5.5.1 The Line Digram

It consists of a number of vertical lines erected on the mid value gives class intervals,
the length of the lines being proportional to the frequency of the respective classes.

5.5.2 The Rectangular Histogram

It consists of rectangle erected on the true class intervals, the aress of these rectangles
are proportional to the frequency of the respective classes. In the case when length of the
classes are equal, the length of the rectangles are porportional to the frequencies of the respective
classes. This diagram is similar to bar diagram, without gaps between bars.

5.5.3 The Frequency Polygon

If the consecutive tops of the line diagram are connected by straight line segments then
the diagram is called a frequency polygon. Tops of the line diagram corresponds to points whose
coordinates are the mid values of the class interval and the corresponding frequencies.

5.5.4 The Frequency Curve

Instead of joining the consecutive tops of the line diagram by straight line segments, if
we join them by means of a smooth, free hand curve then we get a frequency curve.

Problem 6

Draw the line diagram, Histogram, frequency polygram and frequency curve for the
following data.

Marks Number of Students

0-9 3
10-19 5
20-29 10
30-39 20
40-49 45
50-59 32
60-69 26
70-79 15
80-89 4
Total 160
104
 105

5.5.5 Cumulative Distribution

A frequency distribution gives the number of variables that lie in a class interval where as
the cumullative distribution gives the total number of variables that are greater than or less than
a certain value. Accordingly for a frequncy distribution we get two cumulative distributions. They
are known as greater than cumulative distribution and less and than cumulative distribution.

5.5.6 Cumulative Polygon And Ogive

We can represent a cumulative distribution by mean of a graph. If we plot the points

whose coordinates are the end values of the class interval and their corresponding cumulative
frequency and if we connect the points by means of straight line segments we get a Ogive. As
in the case of frequency curve, if we draw a smooth curve joining the points we get a cumulative
frequency curve or Ogive. Thus there are two Ogives for a given frequency distribution.

Problem 7

Draw the two Ogives for the data in Problem 7 and locate the median,

Less Cumulative Greater than Cumulative

(Upper Boundary) Frequency (Lower Boundary) Frequency

9.5 3 0 160

19.5 8 9.5 157

29.5 18 19.5 152

39.5 38 29.5 142

49.5 83 39.5 122

59.5 115 49.5 77

69.5 141 59.5 45

79.5 156 69.5 19

89.5 160 79.5 4

89.5 0
106

5.5.8 Uses of Ogive

The x coordinate of the point of intersection of the less than and greater than Ogive
curves gives the value of the median, a measure of location of the frequency distribution. If only
one Ogive curve is drawn, the x coordinate of the point on the curve whose ordinate is equal to
half of the total frequency gives the values of the median.

(2) The point of inflexion of the Ogive curves gives the value of the mode, another measure
of central tendency. The coordinate of the point on this Ogive curve at which its inclination of the
tangent to the curve is a maximum gives the mode of the distribution.

Check Your Progress II

Fill in the blanks

a. For constructing a graph we generally make use of _____________whereas a diagram

is constructed on a ________________

b. If we draw a perpendicular on the x axis from the point where both less than and more
than curve meet. we get _________________

i) Median, ii) Mode, iii) Mean

 107

5.6 Summary
Statistical data may be displayed pictorially, such as different kinds of diagrams graphs
and maps.

5.7 Check Your Answers

I. a) 10000, b) one, c) T

II. a) Graph paper, Plain paper b) i

5.8 Model Questions

1. Construct a suitable diagram for the following data

Years Export Import

1950-51 100 100

1951-52 124 140

1952-53 162 156

1953-54 167 150

2. Draw the component bar diagram and the percentage bar diagram for the following
data.

Item of expenditure Expenditure in Rupees

Family A Family B

1. Food 700 400

2. Clothing 100 150

3. Education 150 200

4. House Rent 300 250

5. Savings 100 150

3. Draw Pie Diagram for the data in question (2)

108

4. Draw the Histogram, Frequency polygon and frequency curve for the following data.

Income (For Rs.) Number of Individuals

100-149 21

150-199 32

200-249 52

250-249 105

300-349 62

350-399 43

400-449 18

450-499 7

Total 340

5. Draw a histogram for the following frequency distribution.

Age Group: 0-10 10-20 20-30 30-40 40-60

[Link] Persons: 60 40 150 110 110

6. For the folowing frequency distribution draw the Ogives and locate the median.

Class Interval Frequency

90-100 16

100-110 22

110-120 45

120-130 60

130-140 50

140-150 24

150-160 10

7. For the following table gives the mark of students in statistics. Draw the Ogives.

Marks: 0-10 10-20 20-30 30-40 40-50

[Link] Studets: 5 10 10 20 5
 109

LESSON - 6
MEASURES OF CENTRAL TENDENCIES
Structure

6.1 Introduction

6.2 Learning Objectives

6.3 Requisites of Good Average

6.4 Types of Averages

6.5 Mean

6.6 Median

6.7 Mode

6.8 Geometric Mean

6.9 Harmonic Mean

6.10 Summary

6.11 Check Your Answers

6.12 Model Questions

6.1 Introduction
In the lesson we shall study the various measures of central tendency, also known as
average. A statistical average is single number around which the greatest proportion of the data
concentrates. Since it gives the indication of the central tendency of the distribution it is also
called a measure of Central tendency. An average is considered as typical representatives of
the whole data.

6.2 Learning Objectives

After studying this lesson, you should be able to understand:

• the different kinds of average

• the Requisites of a good average

110

• the merits and demerits of different averages

• calcualtion of the averages.

6.3 Requisites of Good Average

A good average should possess the following characteristics

1. It should be well defined.

2. It should be easy to calculate.

3. It should be bassed on all the observations.

4. It should be capable of further algebraic manipulation.

5. It should not be affected by extreme observations.

6. It should not be affected by fluctuations in sampling.

For grouped data midvalues of intervals are used

6.4 Types of Average

The following are the important types of averages.

1. Arithmetic Mean

2. Median

3. Mode

4. Geometric Mean and

5. Harmonic Mean

6.5 Arithmetic Mean (A.M.)

a) Direct Method

Arithmetic mean or simply mean is the commonly used average. We know that if xn are
the n values of an observation then their Arithmetic Mean denoted by x and in given by the
formula.
 111

In the case of frequency distribution f1, f2, ... f are the frequencies corresponding to the
values x1, x2, ...................xn, the A.M. is

b) Short Cut Method to Find A.M.

Let x1, x2, ..................xn be the mid value of the classes and the corresponding frequencies
are f,. f ... f. Let C be the true class interval. We know that

Problem 1

Calculate the mean for the following data

35, 62, 56, 85, 20, 76, 92, 65, 36, 45

35+62 +56 +85 +20 +76 +92 +65 +36 +45
A.M. X =
10

= 572 / 10
= 57.2
112

Problem 2

Calculate the mean for the following frequency distribution

Class Interval Frequency

10-19 2

20-29 8

30-39 25

40-49 36

50-59 52

60-69 45

70-79 35

80-89 12

90-99 2

Solution

Mid x f fx

14.5 2 29.0

24.5 8 196.0

34.5 25 862.0

44.5 36 1692.0

54.5 52 2834.0

64.5 45 2902.0

74.5 38 2831.0

84.5 12 1014.0

94.5 2 189.0

Total 220 12460.0

 113

12460.0
X= = 56.636
220

By Shortcut Method

Take A = 54.5
x1 - A
Mid x f, d= f1d1
C

14.5 2 -4 -8

24.5 8 -3 -24

34.5 25 -2 -50

44.5 36 -1 -36

54.5 52 0 0

64.5 45 1 45

74.5 38 2 76

84.5 12 3 36

94.5 2 4 8

165

-118

47

x
114

Properties of A.M

1. The algebraic sum of deviations of the observations from the arithmetic mean is zero.

Let x1, x2,,.... xn be the n observations and let be their A.M. Then the deviations are

2. If each of the observation are increased by a constant k, then their A.M. is increased
by k.

3. If each of the observations are multiplied by a constant k, then their A.M. is multiplied by
k.

4. If , are the means of two sets of n1 and n2 observations then their combined.

A.M. =

Merits and Demerits of A.M

1. A.M. has a well defined formula

2. The value of A.M. is based on all the observations

3. It is amenable for further manipulations

4. It is easy to calculate

5. It is least affected by fluctuations of sampling.

The greatest disadvantage of the mean is that is affected by extreme (abnormal)

observation. It cannot be determined unless all the classess and frequencies are known. It may
not be the value of any observation. But since A.M. satisfies most of this characteristics of a
good average, it is considered as the best of all average.
 115

Check Your Progress - I

Choose the correct answer

a. The sum of the deviations of individual observations is zero from

i) Mode ii) Median iii) Arithmetic Mean

b. Which average in affected most by extreme observations

i) Mode ii) Arithmetic Mean iii) Geometric Mean

6.6 Median
Median is defined as the value of the middle observations, if the observations are arranged
either in the ascending order or descending order of magnitude. If there are an odd number of
observation then the median is uniquely determined. For example, median of 22, 35, 46, 80, 92
is 46. If there are even number of observations, there are two middle observations, there are two
middle observations, their A.M. is taken as meidan.

For example median of 22,35,46,54,80 and 92 is

46+54 100
= = 50
2 2

So Median is that value of this variate such that there are equal number of observations
above and below that value.

For the case of frequency distribution median is calculated by the formula

N
Median = 1 + /2 - m x c
f

where I is the lower boundary of the median class. i.e., the class in which the cumulative
frequency N/2 occurs, C is the width of the median class, f is the frequency of the median class,
m is the cumulative frequency upto I.

Problem 3

Find the median for the following frequency distribution.

116

Class Interval Number of Students

0-9 17

10-19 31

20-29 36

30-39 159

40-49 97

50-59 73

60-69 35

70-79 35

80-89 11

90-99 6

Solution

end x Cumulative Frequency

9.5 17

19.5 48

29.5 84

39.5 243

49.5 340

59.5 413

69.5 448

79.5 483

89.5 494

99.5 500

N = 500 N/2 =250

It lies between 243 and 350.

Hence Median Class is 39.5-49.5

 117

1 = 39.5, m = 243, N/2 = 250, f = 97

Median

As seen in lesson (1), median can be located graphically from the Ogive curves.

Merits and Demertis of Median

Merits

1. It is easy to understand

2. It can be determined easily

3. It is unaffected by extreme observation

Demerits

1. It is not well defined

2. It is not based on all observations.

3. It cannot be found accurately for grouped distributions

4. It is affected more by fluctuation of sampling than the mean.

5. It is not amenable to further algebraic manipulation.

Check Your Progress - II

a. The positional measures of central tendency in

i) Geometric Mean ii) Mode iii) Median

118

b. The value of median can be determined graphicaly T/F

C. For calculating median it is necessary to arrange the data T/F

6.7 Mode
Mode is defined to be the value of the variable which occurs most frequently. For example
the mode of the set of values 2,3,3,4,6,3. In a frequency distribution of discrete variable, the
mode can be determined easily. For example, if we have a frequency distribution as

x 4 6 8 10 12

f 15 25 12 36 30

Since 10 occurred for a maximum of 36 times mode of the distribution is 10.

In the case of frequency distribution of continuous variable mode is given by the formula

Mode =

where I is the lower boundary of the modal class. i.e., the class for which frequency is
maximum. Here

f1 - the frequency of the modal class.

f0 - the frequency of the class preceding the modal class

f2 - the frequency of the class succeeding the modal class.

C - width of the modal class.

Problem 4

Calculate the mode for the following distribution

Class Limit Frequency

45-49 2

50-54 3

55-59 5

60-64 7
 119

65-69 9

70-74 11

75-79 7

80-84 2

85-89 3

90-94 1

True class Interval Frequency

44.5-49.5 2

49.5-54.5 3

54.5-59.5 5

59.5-64.5 7

64.5-69.5 9

69.5-74.5 11

74.5-79.5 7

79.5-84.5 2

84.5-89.5 3

89.5-94.5 1

Since the maximum frequency 11 occurs for the class interval 69.5-74.5, the modal
class is 69.5-74.5

|= 69.5, f1= 11, f0 = 9, f2 = 7

Mode
120

Merits and Demerits of Mode

Merits

1. Mode is easy to understand

2. Extreme values do not affect the value of the mode.

Demerits

1. It has no definite formula.

2. It is not amenable to further manipulation.

3. It is affect to a greater extent by fluctuations of sampling than the mean

If the frequency distribution is perfectly symmetrical, the three averages mean, medianand
mode will be equal. For moderately asymmetrical distribution an empirical relaion between the
three averages exist. It is given by mean - mode = 3 (mean - median). Given any two of the
averages we can find the third using the above relation.

Check Your Progress - III

a. In a symmetrical distribution mean - median - mode

b. Mode in the highest point of the frequency curve T/F

c. The value of mode can be determined graphically T/F

6.8 Geometric Mean

1. If x1, x2 ................... xn, are n observations, then their geometric mean G.M. is defined as

G.M. = (x1, x2 ......x)1/n = Antilog [1/n [logx1 +log x2 +...logxn]]

2. For the case of frequency distribution

GM = [x1f, x2f..... xnfn] 1/f1 + f2 +.....f n where xi are the mid value of the classes with
frequencies f

Taking logarithm,

log GM. =

=
 121

Problem 5

Find the G.M. for the following observations

56, 35, 60, 78, 22, 40

Solution

GM. = [56 x 35 x 60 x 78 x 22 x 40] 1/6

log G.M. = 1/6 [log56+log35+ log60+ log78 + log22 + log40]

= 1/6 [1.748 + 1.544 +1.778 + 1.892 + 1.342 + 1.602]

= 1/6 [9.906]

= 1.651

GM. = Antilog (1.651)

= 44.7

6.9 Harmonic Mean

Harmonic Mean (H.M) is defined as the reciprocal of the A.M. of the reciprocal of the
observations. If x1, x2

For the case of frequency distribution

X1
122

Check Your Progress - IV

a. The Geometric mean in the __________________ the root of the product of all the
measurements.

b. The harmonic mean in the reciprocal of the arithmetic mean of the reciprocal of the
values T/F

6.10 Summary
In this lesson we have learnt to find the different types of averages.

6.11 Check Your Answers

I. a) iii, b) ii

II. a) iii b) T c) T

III. a) =, b) T c) T

IV. a) n b) T

6.12 Model Questions

1. Compute the arithmetic mean, median and mode for the following distribution of marks
of 500 students in Mathematics.

Scores in Mathematics No. Of students

0-9 17

10-19 31

20-29 36

30-39 159

40-49 97

50-59 73

60-69 35

70-79 35

80-89 11

90-99 6

Total 500

Ans:x=43.1, Median 40.22, Mode 36.15

 123

2. The following frequency distribution relates to the age of first walking of infants in a
locality. Calculate the mean, meidan, and mode.

Age in months No. Of Infants

8-8.9 1

9-9.9 9

10-10.9 20

11-11.9 29

12-12.9 60

13-13.9 32

14-14.9 30

15-15.9 14

16-16.9 8

17-17.9 1

Total 204

X = 12.80; Median = 12.67; Mode = 12.48

3. Find the mean and meidan of the following distribution and determine the mode empirically

Class Frequency Class Frequency

20-40 6 120-140 15

40-60 9 140-160 10

60-80 11 160-180 8

80-100 14 180-200 7

100-120 20

Ans: Mean = 110; Median = 110; Mode = 110.

124

LESSON -7
MEASURES OF DISPERSION
Structure

7.1 Introduction

7.2 Learning Objectives

7.3 Range

7.4 Quartile Deviation

7.5 Mean Deviation

7.6 Variance

7.7 Relative Measures of Dipersion

7.8 Moments

7.9 Skewness

7.10 Kurtosis

7.11 Summary

7.12 Check Your Answers

7.13 Model Questions

7.1 Introduction
An average being a single number which represents a given data does not show as to
how the individual observations are scattered about the average. Two distributions may have
the same number of observations and have the same average. Yet in one distribution the individual
observations may be closely packed around the averages whereas in the other the individual
observations may be widely apart from the measures of central tendency. Such kind of variability
of the individual observations from the central value is known as dispersion or scatterness. In
this lesson we shall consider some important measures of dispersion such as range, Mean
deviation, Standard deviation, Quartile deviation etc.,
 125

7.2 Learning Objectives

After studying this lesson, you should be able

• to calculate mean deviation and standard deviation for grouped and ungrouped data

• to locate median, quartile, deciles and percentile from Ogive curve.

• to calculate the first four moments of a distribution.

• to find out the skewness and kurtories of distributions.

7.3 Range
Range is defined as the difference between the highest and lowest of the given
observations. It can be calcualted very easily. But it depends on only the two extreme
observations, it is not a good measure of dispersion. But the range is to a very useful measures
in industrial engineering work expecially in Statistical Quality Control Work.

7.4 Quartile Deviation

Quartiles are those values of the variate which divide the total frequency into four equal
parts; deciles and percentiles divide it into ten and one hundred equal parts respectively. If the
values of the variate are arranged in ascending order of magnitude then the value of the variate
such that 1/4th of the total observations are less than that value is known as the first quartile or
lower quartile denoted by Q,. Similarly third or upper quartile Q, is th value of the variate such
that 3/4th of the total observations are less than that value.

Since median is the value of the variate such that 1/2 of the total observation are less
than that value, Median correspond to the second quartile Q2. Median also correspond to 5th
decile D5 and the 50th percentile P50. The difference between the upper and lower quartiles i.e.
Q -Q
Q3 - Q1, is called interquantile range and 3 1 / is called the Quartile Deviation. For frequency
2
distribution the quartile are given by

Qi =/ + iN/4-m x c

i = 1,2,3

jth decile, Dj = I + JN/10 - m / f x C j = 1,2...9

126

kth decile, Dk = I + KN/100 - m k = 1,2,....99

x C,
1
where I,n,m,f,c have similar meaning as in the calculation of median.

Problem 1

Find the Quartile deviation, 4th decile and 60th Percentile for the following distribution.

Marks Frequency

1-10 5

11-20 22

21-30 25

31-40 50

41-50 45

51-60 30

61-70 3

180

Solution

end x Cumulative Frequency

10.5 5
20.5 27
Q1
30.5 52
D4
40.5 102
50.2 147
60.5 177
70.5 180
 127

Graphical Location of Quartiles

For to find the partition value graphically, draw the Ogive taking endx on the x axis and
cumulative frequency on the y axis. We have already seen that the value of the median correspond
to the x coordinate of the point of intersection of the Ogives. By drawing only one Ogive alone we
determine the median. Consider the less than Ogive curve. Draw lines parallel to x axis at

The x coordinates of the point of intersection correspond to the value of Q,, Median, D3,
D3 and P45, respectively

Problem 2

The following are the marks of 100 students. Find the values of Q,, M, Q. Graphically

Marks 0-10 10-20 20-30 30-40 40-50 50-60

Frequency 8 20 25 30 12 5
128

Solution

end x Less than Cumulative Frequency

10 8

20 28

30 53

40 83

50 95

60 100

N = 100

N/4 = 25, 3N/4 = 74, N/2 = 50

7.5 Mean Deviation or Average Deviation

Mean deviation about an origin is defined as the mean of the absolute deviation of the
individual values from the given origin. Usually the origin is taken either as the A.M. or the Median
M. If x1, x2, .... xn are n observation, then the mean deviation from

mean =
 129

For the case of the frequency distribution, the mean deviation from mean x is given by

Similarly the mean deviation from median and the mean deviation from the mode can be
obtained.

Problem 3

Calculate the mean deviation about mean and mean deviation about the median for the
following data.

15,25,32,46,80,95,98

Solution

15+25+32 +46 +80+95 +98

The mean =

Mean deviation about mean is

= 1 /7 15-55.8625-55.86 +32-55.86 +46-55.86 +80-55.86 +95-55.86 +98-55.86

= 40.86 + 30.86 +23.86 +9.86 +24.14+39.14 + 42.14

7

= 210.86 / 7

= 30.123

Median = 46

Hence mean deviation about median is

130

Problem 4

Calculate the mean deviation about the mean for the following data

Class 0-10 10-20 20-30 30-40 40-50 50-60 60-70

Frequency 4 10 15 35 52 40 28
 131

7.6 Variance
The mean square deviation of the variate x from any value a is defined as the arithmetic
mean of the square of the deviations of x from a. So the mean square deviations from a is given
by

S2 = 1/N f1 (x1 - a)2

The positive square root of this quantity is called the root mean square deviation from a.
When deviations are taken from the mean x the mean square deviation is called variance and is
denoted by µ2, or 2. So we have

The positive square root of the variance is known as the standard deviation denoted by


132

i.e., the root mean square deviation is least when deviations are measured from the
mean. That is the standard deviation is the least root mean square deviation.

Problem 5

Find the mean and variance of the distribution in which the value of x are 1,2,...,n.

Mean

Mean square deviation from 0 is

 133

Problem 6

The following table gives the monthly wages of workers in a factory. Calculate the standard
deviation.

Monthly wages in Rs. No. of Workers

125-175 2
175-225 22
225-275 19
275-325 14
325-375 3
375-425 4
425-475 6
475-525 1
525-575 1
Total 72

Solution

Let A = 300
134

Theorem

The standard deviation is unaltered by change of origin but its value gets changed by the
change of scale.

Proof

Let a be the origin and scale be h times the original scale then

x2 = h22 where 2, 2 are the variance of x and u respectively. Hence Standard deviation
of x = h times standard deviation of u.

The standard deviation is the best measures of dispersion since

1. It can be calculated easily.

2. It is amenable to algebraic manipulations

3. But it is not suitable in the application of sampling techniques in statistical surveys.

 135

Arithmetic Mean and Standard Deviation of a Combined Set.

If X, and  are the mean and standard deviation of a sample of size n1 x2 and 2 the mean
and standard deviation of another sample of size n2. Then if x and  be the mean and standard
deviation of the combined set n1 + n2 observations then

Proof

Total for the first sample is n1

Total for the second sample is n2 2

Total for the combined set of n1+ n2 observation = n1 + n2 2

Hence Arithmetic Mean of the combined set =

denote summation for the I and II sets respectively.

[~ Algebraic sum of deviation from A.M. = 0]

Similarly, , (x - )2 = n2 22 + n2( 2

+ )2
136

Problem

For any discrete distribution, show that the mean deviation about mean is not greater
than the standard deviation.
 137

We have to show that

Let yi then we have to prove that

i.e., variance of y > 0. which is true.

Problem 7

The means of two samples of sizes 50 and 100 are 54.4 and 50.3 respectively and their
standard deviation are 8 and 7 respectiely. Obtain the mean and standard deviation of the
combined sample of size 150.

Solution

Here n1 = 50, n2 = 100

= 54.4, = 50.3

1 = 8, 2 = 7

Let and be the mean and standard deviationof the combined set
138

7.7 Relative Measures of Dispersion

The measures of dispersion such as the Range, Quartile Deviation, Mean Deviation
standard deviation are called absolute measures of dispersion. There are in the respective unit
of measurements and so they cannot be used to compare dispersion of different groups of
different units. So, for these purposes a pure number independent of the units of measurement
is required. Such a number is called a coefficient of dispersion. These coefficient are known are
relative measures of dispersion.

7.7.1 Co-efficient of Range

Maximum - Minimum
Maximum + Minimum

7.7.2 Co-efficient of Quartile Deviation

Q3 - Q1
Q3 + Q1

7.7.3 Co-efficient of mean deviation

7.7.4 Co-efficient of variation

The measures of dispersion discussed above are absolute in nature for they are
expressed in the unit in which the data is collected. For example, the standard deviation of the
heights of students may be in inches where as the standard deviation of weights may be in Kg.
 139

So comparison of dispersion by these measures are not possible. So we are interested to have
a relative measure of dispersion, which will be a number free from unit of measurement.
Coefficient of variation (C.V) is the most important relative measure of dispersion

The co-efficient of variation is used as a measured of consistency. Any data with the
least value of C.V is the most consistant

Problem 8

The following are the wages of workers in two factories. Find which factory's wage is
more variable.

Weekly wages in Rs. Number of workers

Family A Family B
1-100 25 15
101-200 30 20
201-300 22 30
301-400 20 35
401-500 15 20
112 120

Solution

For Factory A
140

For Factory B
 141

Since the coefficient of variation for factory B is less than the coefficient of variation for
factory A. Factory A's wage is more variable.

Check Your Progress - I

Say True or False

a. Range is the best measure of dispersion T/F

b. There is no difference between variance and co-efficient of variation T/F

c. The quartile deviation is more suitable in case of open-end distributions T/F

d. Mean deviation is least when deviations are taken from median T\F

e. The variance is equal to square of standard deviation T/F

142

7.8 Moments
The rth moment about any point a, denoted by µ is given by µr= 1/N f1 (x1 - a)r N........(1)

Thus the rth moment about 'a' can be regarded as the mean of the variate (X-a)".

The rth moment about the mean X, denoted by µ, is given by

Putting r 1 in (1) and (2) we get

Moments about the mean interms of moment about any other point.

Putting X, = x,-a and d = X, a we have

 143

Since d = µ1' we get

µ1 = µ1'-fc1µ1 µr-1' + fC2 (µt')2 µr-2 + .... (-1)rµ1

Putting r = 2,3,4 we get

µ2 = µ2' - 2(µ1')2 + 2(µ1 ')2 = µ2' - µ1')2

µ3 = µ3' - 3µ1' µ2' + 3(µ1')2 µ1' – (µ1')3

= µ3' - 3µ1 µ2' + 2(µ1')3

µ4 = µ4' - 4µ1'µ3' + 6(µ1')2 µ2' - 4(µ1')3µ'1 + (µ1')

= µ4-4µ1 µ2' + 6(µ1')2 µ2' - 3(µ1')4

Conversely we have the result

µr' = µr + fc1µr-1d + rc2µr-2d2 + ........... rc1µ2dr-2 + df

Effect of Change of Origin and Scale

Let u be a variate given by u= x - a / h

Then x = a+uh :. = a+hu

x- =h(u-u) then

The rth moment of the variate x = h' times the corresponding moment of the variate u.

Take a = and h = , we see that the distribution of u has zero mean and unit variance.

Problem 9

Calculate the first moment about the mean for the following data.
144

x 1 2 3 4 5 6 7 8 9

f 1 6 13 25 30 22 9 5 2

X1 -a
X1 f1 d1 = f1d1 f1d12 f1d13 f1d14
C

1 1 -4 -4 16 -64 256

2 6 -3 -18 54 -162 456

3 13 -2 -26 52 -104 208

4 25 -1 -25 25 -25 25

5 25 -1 -25 25 -25 25

6 30 0 0 0 0 0

7 22 1 22 22 22 22

8 9 2 18 36 72 144

9 2 4 8 32 128 128

113 -10 282 2 1674

 145

Sheppard's Corrections to Moments of Grouped Frequency

In grouped frequency distribution we had assumed that the frequencies are concentrated
at the midpoints of the class interval. This assumption is nearly true if the distribution is symmetrical
or slightly symmetrical and the class intervals are not greater than 1/20 the of the range. But it
is not in general true. So an error called grouping error creep in the calculation of moments.
W.F. sheppard had proved that if (1) the frequency distribution is continuous and (2) frequency
tapers off to zero in both directions.

Then

µ2 (corrected) = µ2-h2 / 12

µ3 (corrected) - µ3
146

µ4 (corrected) = µ4 - (1/2) h2 µ2+ (7/240) h4

where h is the width of the class interval. These correction are known as Shappard's
corrections.

Pearson's  and  coefficients: Karl Pearson defined the following four coefficients
based upon the first four moments about mean. They are

These coefficients are pure numbers independent of unit of measurement.

7.9 Skewness
Skewness means lack of symmetry. Skewness indicates whether the frequency curve
is inclined more to one side than to the other. If the mean of distribution is greater than the mode,
then the skewness is said to be positive. If the mode of the distribution is greater than the mean,
then the skewness is said to be negative. The purpose of measuring skewness is to estimate
the direction in which and also the extent by which the distribution is distored from symmetry.

7.9.1 Karl Pearson's Measure for Skewness

Skewness

But since Mean - Mode

We have Skewness = 3 (Mean - Mode) / S.D

this is a number free from units of measurement. It is zero for symmetric distribution
since for distribution mean, median and mode coincide.

7.9.2 Another measure for skewness called Bowley's coefficient

Skewness = Q1 + Q3 - 2M / Q3 - Q1

7.9.3 Measure of Skewness based on Moments

The measure of skewness is obtained by making use of the third moment about the
mean. When the method of moment in applied,  is used as a relative measure of skewness. 
is defined as
 147

The value of B, shall be zero for a perfectly symmetrical series. The greater the value of B, the
more skewed the series. Instead of ?,, Karl Pearson suggested y, to be used as a measure of
skewness. He defined Y1 as

For a symmetrical distribution  will be zero. If the value of  is positive the distribution
is positively skewed and if it is negative the distribution is negatively skewed.

So for symmetric distribution = 0. So , can also be taken as a measure for

skewness. The sign of is the same as the sign of 3.

148

7.10 Kurtosis
For some distribution the peak around the mode may be sharp. For some other it may
be flat. Kurtosis measure the nature of the peak around the mode. It is measured by B2. If the
peak of the frequency curve near the mode is flat the curve is said to be platy kurtic. If the peak
is normal, the curve is set to be mesokurtic. If the Peak of a frequency curve is sharp, then the
curve is said to be leptokurtic. The frequency curve is platykurtic, mesokurtic or leptokurtic
according as  < 3, or 3 or > 3

Where  = 

 = 4th Moment

 = 2nd Moment

1. Platykurtic, 2. Mesokuric, 3. Leptokurtic

Note: Percentile range / Q.D. also measures kurtosis

Check Your Progress - II

Choose the Correct Answer

a. When coefficient of skewness is O is the distribution is

i) J. Shaped ii) U-shaped

iii) Symmetrical iv) L-Shaped

 149

b. If the frequency distribution is positively skewed the mean of the distribution is

i) Greater than mode ii) Lesser than mode

iii) Equal to the mode

7.11 Summary
In this lesson we have learnt the measures of dispersion, relative measures of dispersion,
to calculate the first four moments of a distribution and also to find out the skewness, kurtosis of
a distribution.

7.12 Check Your Answers

I. a) F, b) F, c) T, d) T, e) T

II. a) iii b) i

7.13 Model Questions

1. Calculate the average deviation and the standard deviation for the following distribution.

Height in Inches Frequency Height in Inches Frequency

56-58 1 66-68 134

58-60 18 68-70 28

60-62 50 70-72 7

62-64 142 72-74 1

64-66 219

Total 600

Ans: Average Deviation = 1.8, S.D. = 2.36

2. The mean height and standard deviation of height of sample of 400 individuals are found
to be 65.4 "and 2.31" respectively. In another sample of 600 individual they were found to
be 66.6" and 2.34" respectively. Find the mean and standard deviation of the combined
of 1000 individuals.
150

3. The data below gives the run taken by two cricketers A and B in 27 innings. Find out who
scored more on average and who is more consistent.

A-301, 214, 199, 174, 151, 139, 123, 49, 80, 74, 51, 49, 41, 39, 37, 31, 30, 29, 24, 21, 19,
17, 17, 15, 12, 8, 5,

B-291, 276, 237, 190, 125, 94, 89, 74, 65, 57, 51, 47, 45, 42, 30, 29, 25, 14, 9, 8, 7, 7, 6,
5, 2, 0, 1

[Ans: x for A = 72.2 x for B = 68 C.V. for A= 104.84 C.V. for B = 111.77]

A scored more on average and he is more consistent.

4. Find the standard deviation and coefficient of variation for the following distribution.

Class 20-25 25-30 30-35 35-40 40-45 45-50

Frequency 170 120 80 45 30 35

Ans: S.D. = 17.4 C.V. = 11.76

5. Find the standard deviation and coefficient variation from the following data.

Age Group [Link] Persons

20-25 50

25-30 70

30-35 100

35-40 180

40-45 150

45-50 120

50-55 70

55-60 60

Ans. X = 40.31  = 9.35, C.V.23.19%

 151

6. Calculate Karl Pearson's coefficient of skewness of the following.

Marks [Link] Students

0-10 30
10-20 40
20-30 50
30-40 60
40-50 26
50-60 162
60-70 132

7. Find the quartile deviation and quartile coefficient of skewness for the data.

Income 0-10 10-20 20-30 30-40 40-50 50-60 60-70

[Link] Person 22 240 350 410 220 100 35

8. For the following distribution of marks of 100 students in mathematics and the various
measures of skewness

Scores 0-9 10-19 20-29 30-39 40-49 50-59

[Link] Students 4 6 13 26 20 14

Scores 60-69 70-79 80-89 90-99 Total

No. of Students 8 5 3 1 100

Ans : Pearsons Measure for Skewness = 0.35

Quartite Measures for Skeweness = 0.17

Kurusis based on moments = 3.02

152

LESSON - 8
THEORY OF PROBABILITY
Structure

8.1 Introduction

8.2 Learning Objectives

8.3 Definitions of Various Terms

8.4 Classical Definition of Probability

8.5 Random Experiment

8.6 Independent Events

8.7 Random Variable and Distribution Functions

8.8 Continous Random Variable

8.9 Summary

8.10 Check Your Answers

8.11 Model Questions

8.1 Introduction
The first foundation of the Mathematical theory of Probability was laid down in the mid-
seventeenth century by two French Mathematician B. Pascal and P. Fermat while solving
problems of game of chance. Probability today has become one of the fundamental tools of
statistics.

8.2 Learning Objectives

After studying this lesson, you should be able to:

• understand various definitions of probability and their limitations

• prove theorem in probability

• understand definition of Conditional Probability, Baye's theorem and its applications

• understand the terms random variable, Probability density function, Distribution function.

• calculate moments for discrete and continuous random variable.

 153

8.3 Definitions of Various Terms

We shall discuss about the different terms of the probability.

Trial and Event

Consider an experiment which though repeated under essentially identical conditions,

may not given unique results but may result in any one of the several possible outcomes. The
experiment is known as trial and outcomes are known as events or causes. For example,
tossing a coin in a trail and getting a 'Head' or 'Tail' is an event.

Exhaustic Events

The total number of possible outcomes in any trial is known as exhaustive events. For
example in tossing a coin there are two exhaustive cases namely head and tail, if we assume
that the coin cannot stand on its edge.

Favourable Events

The number of cases favourable to an event in a trail is the number of outcomes which
entail the happening of the event. For example, in the tossing of a die, the number of cases
favourable for the appearance of a multiple of 3 are two namely 3 and 6

Mutually Exclusive Events

Events are said to be mutually exclusive or incompatible if the happening of any one of
their precludes the happening of all others. i.e., no two or more of them cannot happen
simultaneously in the same trial. In tossing a coin, getting a head and getting a tail are mutually
exculsive events.

Equally Likely Events

Events are said to be equally likely when we have no reason to expect any one than the
other. For example, in tossing a uniform coin, getting a head or a tail are equally likely events.

8.4 Classical Definition of Probability

If a trail results in an exhaustive, mutually exclusive and equally likely cases and m of
them are favourable to the happening of an events, E, then the probability 'p' of the happening is
given by

P=
154

Since the number of cases favourable to the non-happening of the event E are n-m the
probability of the E not happening is given by

Therefore, p+q = 1 p and q are non negative and cannot exceed unity i.e., 0 < P < 1,
0 < 9 < 1.

If the probability for an event E is 1, the event is called a Sure or Certain Event and if the
Probability =0, then the event is called an impossible event.

Consider the experiment of tossing a fair coin. There are two equally likely cases, head
and tail and hence the probability for a head is 1/2. Similarly if we roll a die, there are 6 equally
likely exhaustive cases. Suppose we require the probability of getting an odd number, since
there are 3 odd number viz 1,3,5, the number of cases favourable to the event is 3. Hence the
probability of getting an odd number in throwing a die is 3/6 = 1/2.

8.4.1 Limitation of the Classical Definitions

When the events are not equally likely and also when the exhaustive number of cases in
a trail is infinite, this definition fails.

8.4.2 Statistical Probability

If a trail is repeated a number of times under esstially homogeneous and identical

conditions, then the limiting value of the ratio of the number of times the events happens to the
number of trials, as the number of trails becomes infinitely large, is called the probability of
happening of the event provided the limit exist and is finite.

If in n trails an event E happen m times, then the probability p of the happening of E is

given by

Problem 1

What is the probability that a leap year selected at random contains 53 sundays?
 155

Solution

In a leap year which contains 366 days. There are 52 weeks and 2 days. The 52 weeks
contains 52 sundays.

The other two days may be of any one of the following combinations.

1. Sunday and Monday

2. Monday and Tuesday

3. Tuesday and Wednesday

4. Wednesday and Thursday

5. Thursday and Friday

6. Friday and Saturday

7. Saturday and Sunday

For to have 53 sundays, we should have on more sunday from the remaining two days.

The favourable ways for a sunday is 2. Toal ways = 7

Therefore, Probability for 53 sunday is 2/7

Problem 2

From a well shuffled pack of 52 cards. Two cards are drawn at random. Find the probability
for them to be king.

Solution

There are 4 kings in a pack of 52 cards. We can get 2 kings from the 4 in 4C2 =
4x3
/ 1 x 2 = 6. We can draw any two cards from the pack of 52 cards in 52C2 ways = 1326.

Therefore, Probability for drawing two kings: = 6 / 1326 = 1 / 221

8.4.3 Additive Law of Probability

The probability that one of several mutually exclusive events A,, A,...A will happen is the
sum of the probabilities of the separate events.

i.e., P(A1 + A2+.......................... A) = P(A1) + P(A2) +.......................... P(An)

156

Proof

Let N be the total number of exhaustive, mutually exclusive and equally likely cases and
out of these let m1, m2, ....... mn be respectively favourable to the events A1, A2 .......................An.
Since the n events are mutually exclusive these m1, m2,..........mn cases are distinct and non-
overlapping. Hence the total number of cases which are favourable to either A1 or A2.................or
A1 is m1 + m2 + ..........................mn

Hence P(A1 + A2 + ........................An)

Problem 3

From a well shuffled pack of 52 cards, two cards were drawn at random. Find the
probability that they are both clubs or both diamonds.

Solution

Two cards can be selected out of 52 in 52C2 =

Total number of Clubs is 13

Number of ways of selecting two clubs is 13C2 =

Therefore, probability of drawing 2 cards both of which are clubs is 78/1326 = 1/17.
Similarly, the probability of drawing 2 cards both of which are diamonds is 1/17.

Since the events are mutually exclusive, the probability for two clubs or two diamonds is
1/17 + 1/17 + 1/17

Problem 4

An urn contains 3 red and 4 black balls. Two balls are drawn at random. Find the probability
that the balls are of (1) different colours (2) black color (3) red colour.
 157

Solution

Total number of balls in the urns is 7

7x6
Any two balls can be selected in 7C2 = = 21
1x2

i) For the balls to be of different colours one should be red and the other should be black.
One red ball can be selected in 3C, = 3 ways. One black ball can be selected in 4C1 = 4
ways. So one red ball and one black ball can be selected in 4 x 3 = 12 ways.

Hence the probability for one red and one black ball is 12/21 = 4/7
4x3
ii) There are 4 black balls. Any two black balls can be selected in 4C2 = =6 ways.
1x2

Number of favourable cases is 6. Total number of cases is 21. Hence the probability
for 2 black ball = 6/21 = 2/7
3x2
iii) There are 3 red balls. Any two balls can be selected in 3C2 ways = = 3 ways.
1x2

The total number of cases is 21.

Therefore, Probability for 3 red balls is 3/21 = 1/7

8.5 Random Experiment

An experiment in which chance plays an important role in the outcomes is called random
experiment. i.e., the given condition do not determine completely the outcomes. This kind of
experiment is also known as non-deterministic or probabilistic or stochastic experiment. For
example in rolling an unbiased die one cannot predict with certainly which face will turn up. Each
performance of a random experiment is called a trail. The result of a trail in a random experiment
is called an outcome or a sample point. The totality of all possible outcomes (i.e., sample points)
of a random experiment is called a sample space. For eg. in tossing a fair coin, the sample
space consists of the two points H, T.

A sample space is called finite (infinite) sample space if the number of elements in S is
finite (infinite)

Any subset A of the sample space is known as an event.

158

Consider the experiment of tossing 2 fair coins then the sample space S = {(H,H), (H,T),
(T, H), (T, T)}

If A= (H,H), then A is an event that head appears in both the coins.

Let E be the experiment of the tossing three coins at a time. Then the sample space S is
given by

S = {H, T} x {H, T} x {H, T}

= {HH, HT, TH, TT} x {H, T}

= {HHH, HHT, HTH, HTT, THH, THT, TTH, TTT,}

= { ..................} say

If E1 is the event that the number of heads exceeds the number of tails 1, E2 is the event
of getting two heads and E3 the event of getting 'head' in the first trail then these are represented
by the following sets

E1 = { }

E2 = {}

E3 = { }

8.5.1 Algebra of Events

For events A, B, C

1. A B = {S:  A or  B}

2. A  B = {S:  A or  B}

3. Ac (Complement of A) = { S:  A}

4. A - B = {S:  A but  B}

5. A  B for every  A,  B

6. B A A B

 159

7. A = B if and only if A  B and B A

8. A and B disjoint (mutually exclusive) = A  B = 

8.5.2 Axiomatic Development of Probability Theory

In the axiomatic development of probability theory, Probability is defined as a function

defined as events (subset of S) i.e., it is a rule which associate with every event A a real number
P(A) which satisfies the following axions.

Axiom 1

P(A) > 0 i.e, Probability of every events is non-negative.

Axiom 2

P(s) = 1 i.e., Probability of a certain event is unity

Axiom 3

A1, A2, ...... are a countable number of subevents of S such that

A1  A2,.... = S and Ai  Aj =  (iJ)

P(A1 A2,...) = P(A1) + P(A2) .....

i.e., the probability of a union of disjoint events is the sum of the probabilities of the
events themselves.

Theorem 1

Probability of an impossible event is zero. i.e., P () = 0

Proof

Impossible event contain no sample point and hence the certain event S and the in
possible event are mutually exclusive.

Hence S  =S

P (S E ) = P(S)
160

i.e., P(S) + P() = P(S) [From Axiom 3]

P()= 0

Theorem 2

P(A) = 1-P(A)

Proof

Theorem 3

8.5.3 Laws of Addition of Probabilities

If A and B are any two events, then

P(A B) = P(A) + P(B) - P (A B)

 161

Proof

8.5.4 Theorem of Compound Probability

For two events A and B

where P(B/A) represent the conditional probability of occurrence of B, given that the
event A had already happened and P(A/B) is the conditional probability of A given that B had
already taken place.

Proof

Suppose the sample space contains N occurances of which NA occurances belong to

the event A and NB. occurences belong to the event B. Let NAB be the number of occurences
favourable to the compound event A  B.

Then the unconditional probabilities are given by

162

Now, the conditional probability P(A/B) refers to the sample space if no Occurences out
of which nNR occurences pertain to the occurences of A, that is when B has already happened.

...

Thus the conditional probabilities P(B/A) and P(A/B) are defined if and only if P(A) #0 and
P(B) = 0 respectivesly.

Note that P(A/B) < P(A) using mathematical induction we can extend the result to any n
events.

Proof

For any two events A, and A, we have

P(A1  A2) = P(A1) P(A2/A1) ie., the result (I) is true for n = 2.

Let us suppose that the result is true for n=m

i.e., P(A1  A2, ...........  Am) = P(A1) P(A2/A1).... P(Am /A1 A2.....Am-1)

Now, P(A1  A2......  Am+1) = P(mA1)  Am+1

P(mA1) P(Am+1/ mA1) using (1)

= P(A1) P(A2/A1) P(Am/A1  A2 .............. Am-1) x P(Am+1 / m A1) using (2)
 163

Thus the result of the theorem is true for n = m+1 also. We have proved that the result is
true for n = 2. Hence it is true for all positive integral values of n.

8.6 Independent Events

An event B is said to be independent of event A, if the conditional probability of B given A,
ie., P(B/A) is equal to the unconditional probability of B.

ie., P(A/B) = P(B) Since P(A B) P(B/A).P(A) and Since P(B/A) = P(B) when B is
independent of A, we have P(A  B) = P(A) P(B)

Problem 5

A and B throw alternatively with a pair of dice. One who first throw a total of 9 wins. Find
their respective chances of winning if A starts the game.

Solution

Let E, and E, denote the events A and B throwing a total of 9 with a pair of dice.

Then

P(E1) = P(E2) = 4/36 = 1/9 and

If A starts the game, he can first throw a total of 9 in the following mutually disjoint ways.

1) E, happens (2) E1  E2  E1 happens (3) happens and so on.

Hence by addition theorem of probability the probability P that A first to win is given by

by compound probability theorem

Probability for B to win = 1-P = 1 - (9/17) = 8/17.

164

Problem 6

From a well shuffled pack of 52 cards are dealt one by one until an ace appears. Show
that the probability that exactly n cards are dealt before the first ace appears is
4(51-n) (50-n) (49-n)
[Link]

Solution

Probability for an ace to appear in the first draw is 4C1 / 52C1 = 4/52

So the probability for an ace not to appear in the first draw

1 - 4/52 = 48/52

Probability that the first ace not to appear in the second draw is

1 - 4/51 = 47/51

Probability that the first ace not appear in the third draw is

1- 4/50 = 46/50

Probability that the first ace not to appear in the (n-i)th draw is

Probability that the first ace not to appear in the nth draw is

Probability that the an ace to appear in the (n+1)th draw is

4/52 - n

So the probability for the first ace to appear at the (n+1)th draw is
 165

8.6.1 Baye's Theorem

If E1, E2 ............... En are mutually disjoint events with P(E1) 0 (i=1,2,....n) then for any

arbitarary event A which is a subset of E1 such that P(A) > 0.

We have P(E,/A) =

Proof

Since we have

By distributive law

Since (i = 1,2,.......n)are mutually disjoint events we have by addition theorem

of probability.

Problem 7

There are two identical boxes containing respectively 4 white and 3 red balls, 3 white
and 7 red balls. A box is chosen at random and a ball is drawn from it. If the ball is white what is
the probability that it is from first box.
166

Solution

Let E1, E2 be the events of selecting I and II boxes respectively. Since the boxes are
identical P(E) = P(E2) = 1/2. Let A be the selecting a white ball. Then we require P(E1/A).

By Baye's Theorem P(E,/A) =

P(A/E1) = Probability of getting a white ball from the I urn = 4/7

P(A/E2) = Probability of getting a white ball from the Il urn = 3/10

Hence P(E1/A) =

8.7 Random Variable and Distribution Functions

A real number connnected with the outcome of a random experimental is called a random
variable. Consider the experiment of tossing 2 coins at a time we may take the random variable
as the number of heads drawn. Then we have

Outcome HH HT TH TT

Value of x 2 1 1 0

Thus to each outcome w, there is a real number x(). Hence X() is a random variable.

8.7.1 Definition

Let E be an experiment and S be a sample space associates with the experiment. A real
valued function X which assigns to every element of S only one real number x(8) is called a
random variable. It may be possible that S has elements which are themselves real number. In
that case X() = 

A random variable which can assume only a discrete set of real numbers is called a
discrete random variable.
 167

8.7.2 Discrete Probability Distribution

Let x be a discrete random variable. Let the possible value which it assume be x1, x2,
..........xn.

Suppose P(x1 = x1) = p(x1), P(X = x2) = p(x2).

Then the function P(x) is called the probability function or Probability mass function, if it
satisfies the following conditions.

i. P(x) > 0 for every x

ii. P(X1) = 1

The collection of pairs [x1, P(x)] for i = 1, 2, ...... n is called the probability distribution of X.

Problem 8

Consider the experiment of tossing a coin twice. Let X represent the number of heads
which come up. The outcome of the experiment are HH, HT, TH, TT. The values taken by X are

X = 2, 1, 1, 0

and the probability are 1/4, 1/4, 1/4, 1/4

P (X=0) = P {TT} = 1/4

P (X=1) = P {HT} + P {TH} = 1/4 + 1/4 = 1/2

P (X=2)= P {HH} = 1/4

The Probability function of X is

X 0 1 2

P(X = x) 1/4 1/2 1/4

8.7.3 Discrete Distribution Function

The Function F(x) which gives the cumulative probability upto a specified value of the
random variable X is known as the Distribution function of X or the Cumulative distribution of X.
168

If X takes only a finite number of values x1, x2, .................xn. then the distribution function
of X is given by

Fig 8.2 Discrete Probability Density Grap

Problem 9

A random variable X has the following probability function

X 0 1 2 3 4 5 6 7 8

P(X) a 3a 5a 7a 9a 11a 13a 15a 17a

i) Determine the value of a

ii) Find P (X<3), P(X>3), P(0<X<5)

iii) What is the smallest value of x for which P(X<x) > 0.5 and find out the distribution function.
 169

Solution

The distribution function F(x) is as follows.

X 0 1 2 3 4 5 6 7 8

F(x) 1/81 4/81 9/81 16/81 25/81 36/81 49/81 64/81 81/81

Then P (X < x) > 0.5?x=7

Problem 10

From a lot of 25 items containing 5 defective a sample of 4 items was drawn at random
(i) with out replacement (ii) with replacement. If x denote the number of defectives, find the
probability distribution of X.

Solution

We can draw any 4 item from the 25 items in 25C4 ways. The number of ways in which
x defectives can be chosen out 4 of 5 is 5Cx and the number of ways of selection 4- x non
defective from the remaining 20 items is 20C4-x

Hence P(X=x): =

ii) When the items selects are replaced the number of ways of selecting x defective =
4
Cx. The probability of choosing one defective out of 5 is 1/5. So the probability of choosing x
defectives is (1/5). The probability of not choosing a defective is 4/5. The probability of not
choosing 4-x non defective items is (4/5)4-x. Hence the probability of choosing x defective and 4-
x non defective items is P(X=x) 4Cx (1/5)x (4/5)4-x
170

8.8 Continuous Random Variable

A random variable x is said to be continuous if it takes all possible values between certain
limits.

8.8.1 Definition Probability Density Function

Consider a small interval [x-dx/2, x+dx/2] of length dx round the point x. Let f(x) be a
continuous function of x so that f(x) dx represents the probability that X falls in the infinite interval
[x-dx/2, x+dx/2]. Symbolically P[x-dx/2 < x < x + dx/2] ˜ f(x)dx.

In the fig 8.3 f(x)dx represent the area bounded by the curve y = f(x) x axis and the
ordinates at the point x-dx/2 and x+dx/2. This function f(x) is known as the probability density
function or simply the density function of the random variable and is usually abbreviated as p.d.f.
The curve y = f(x) is known as the probability density curve. The probability for variate to lie in the
interval dx is f(x)dx. Hence the probability for a variate to fall in the finite

interval f(x)dx. Since the total probability is unity we have f(x) dx = 1.

Fig 8.3 Probability density curve continuous random variable

 171

The probability density function of a random variable X satisfies the following properties.

Then

F(x) is called the Cumulative distribution function of x or simply the distribution function.
The distribution function satisfies the following properties

ii) Since for the continuous random variable the probability at a particular point is
always zero.

P(a<x<b) = P(a =x=b) = F(b) - F(a)

iii) Since f(x) = 0, F(x) is a non decreasing function of x

iv) If f(x) is a continuous function F'(x) = f(x)

172

Problem 11

A continuous random variable X follows the probability law f(x) = Ax2, 0 = x = 10. Determine
A and find the probability that

a) X lies between 0.2 and 0.5

b) X<3

c) 1/4 <X < 1/2

d) X > 3/4

Solution
 173

Problem 12

The probability distribution function of a random variable X is

Compute the cumulative distribution function of X.

Solution

For any x in the range x = 0,

174

8.8.2 Measures of Central Tendency for Continuous Probability Distribution

8.8.3 Quartiles And Deciles

Q, and Q, are given by the equations

Mode being the value of x which f(x) is maximum, Mode is given by f' (x) = 0 and f'(x)<0.

Problem 13

Calculate the standard deviation and mean deviation from mean if the frequency function
f(x) is
 175

Solution

Check Your Progress

Fill in the blanks

a. Probability ranges from _______________ to ______________

b. Probability theory had its origin in games

C. (i) 5C2 = ______________ (i) 5P2 = ______________

d. If two events A and B are independent the probability that they will both occur is given by
_______________
176

8.9 Summary
In this lesson, we have learnt, the probability theory, probability density function and
distribution function.

8.10 Check Your Answers

I a) 0,1 b) Gambling c) 10, 20 d) P(A) * P(B)

8.11 Model Questions

1. An urn contain 3 red, 4 white and 5 blue balls. Another urn contains 5 red, 6 white and 7
blue balls. One ball is selected from each urn. What is the probability that

i. Both will be white

ii. Both will be same colour

iii. Neither will be red

2. A and B toss a coin alternately on the understanding that the first who get a head wins. If
A stars, show that their respective chances of writing are 2/3 and 1/3.

3. In a bolt factory machines A, B and C manufacturing respectively 25%, 35% and 40% of
the total out put 5, 4, 2 percents are defective bolts. A bolt is drawn at random from the
product and is found to be defective. What are the probabilities that it was manufacured
by machine A, B and C

Ans: 25/69, 28/69 and 16/69

4. A random variable has the following probability distribution

x 0 1 2 3 4 5 6 7

P(x) 0 k 2k 2k 3k k2 2k2 7k2+k

Evaluate P(x<5), P(o< x <5)

5. Define distribution functions and write down its properties.

6. A random variable X has the probability density functions

f(x) = 6x (1-x) for 0 < x < 1 and otherwise. Find the Mean, Mode, Median.
 177

LESSON - 9
MATHEMATICAL EXPECTATION
Structure

9.1 Introduction

9.2 Learning Objectives

9.3 Central Moments

9.4 Conditional Expection

9.5 Expection of a Linear Combination of Random Variable

9.6 Expectation of a Continuous Random Variables

9.7 Moment Generating Function

9.8 Characteristics Function

9.9 Summary

9.10 Check Your Answers

9.11 Model Questions

9.1 Introduction
If x1, x2......xn, are the values of a discrete random variable X with the corresponding

probabilities p1, p2............pn then xnpn is known as Mathematical expectation of the random

variable. Provided xp, converges absolutely. It is denoted by E(X).

E(X) = xnPn

If X is a continuous random variable with p.d.f. f(x) then the expection of X is defined as
E(X) =  x f(x) dx.
178

9.2 Learning Objectives

After studying this lesson, you should be able to

• understand the terms mathematical expectation, conditional expection, moment

generating function and characteristic function

• understand the theorems on mathematical expection

• to do problems in mathematical expectation

9.3 Central Moments

If x1, x2.......xn, are the values of discrete random variable X with the corresponding
probabilities p1, p2.........pn, and  (X) is a function of X which takes the value  (xn) when X takes
the value xn, Then the expectation of  (X)

E[(X)] = (xn)pn

If X is a continuous random variable with the density function f(x) then

 179

The other central moments can be obtained either by substituting different values of r in
, or obtained from the non-central moments.

Problem 1

A continuous random variable X follows the probability density function Kx2 for 0 < x < 10.
Determine the constant K. Find 1 and 2.

Solution
180

Problem 2

A coin is tossed untill a head appear. What is the expectation of the number of tosses?

Solution

Let the head appear only in the nth trial. In the previous (n-1) trials we should get tail and
in the nth trail we should get a head hence the probability for getting the head in the nth trail
(1/2)n-1 (1/2) = (1/2)n

E(n) = 1.(1/2) + 2.(1/22) +..............+ n.(1/2n) + ....

 181

Theorem 1

The expectation of the sum of two random variables is equal to the sum of their
expectations i.e., if X and Y are the random variable the E (X+Y) = E(X) + E(Y)

Proof

Let X assumes the values x1, x2................. xn with corresponding probabilities p1, p2 ....pm
and let Y assumes the value y1, y2 ... yn with corresponding probabilities p'1, P'2 ....... P'n. Then
X + Y is a stochastic variable which can assume any one of the mn values x1 + yj (i = 1, 2.....mj
= 1,2...n). If pij denotes the probability for X assuming a value x1 and simultaneously Y assuming
a value yj, then the probability for X assuming a value x1 is

and the probability for Y assuming a value y, is

182

Note

The above result can be extended to any number of discrete random variables.

i.e., E(X+Y+Z....) = E(X) + E(Y) + E(Z) +.....

Theorem 2

The expectation of the product of two independent random variable is equal to the product
of their expectations. i.e., If X and Y are two independent random variables then E(XY) = E(X)
E(Y)

Proof

Let X assumes the values x1, x2,..... xm with corresponding probability p1,p2, ...pm and Y
assumes the values y1, y2.... yn with corresponding probabilities p'1, p'2, ... p'n the XY can take
any of the mn x1y1, i = 1, 2....m; j = 1,2...n

If p, denotes the probability of X assuming the value x, and simultaneously Y assuming a

value y1, then as X and Y are independent

pij = pi, pj

By definition E(XY) =

The result can be extended to any number of independent random variables. E(XYZ...) =
E(X) E(Y) E(Z).....

9.4 Conditional Expectation

Let X be a random variable which takes the value x, with probability p(y) subject to the
occurence of the event Y.
 183

Then p1(y) = p(x1/y)

The conditional expectation of X given Y is defined as

9.4.1 Variance

The variance of distribution is

Theorem 3

If X is a random variable and 'a' is a constant, them

i) E(a (X)) = a E[(X)]

ii) E((X) + a) = E [(X)] + a where (X) is any function of X

Proof

If X can take the possible value x1, x2 ........xn with respective probabilities p1, p2...pn then
by definition.
184

i) E[a(X)] = a(x1)p, = a (x1)p1 = a E[(X)]

ii) E[a(x)+a] = [(x1)+a] p1 = (x1)p1 + a p1 = E[(x)] + a since p1 = 1

Cor

i) If (X) = x, then E(ax) = aE(X) and E(X+a) = E(X) + a.

ii) If  (X) = 1, then E(a) = a

Theorem 4

If X is a random variable and a b are constant then E(aX + b) = aE(X) + b

By definition, E(aX+b) =

Taking a = 1, b = -x = -E(x), we have

E(x - X) = 0

Theorem 5

If X is random variable, then Var (aX+b) = a2 Var(X) where a, b are constant

Proof

Let Y = ax + b, then E(Y) = aE(X) + b

 185

~ Y- E(Y) = a[X - E(X)]

~ [Y-E(Y)]2 = a2(X-E(X))2

~ E[Y-E(Y)]2 = a2E[X-E(X)]2

i.e., Var Y = a2 Var X

i.e., Var (ax + b) = a2 Var X

Cor

i) If b= 0 then Var (aX) = a2 Var X

ii) If a = 0 then Var (b) = 0

iii) If a = 1 then Var (X+b) = Var X

9.5 Expectation of a Linear Combination of Random Variable

Let X1, X2 .... Xn be any n random variables and if a1, a2......an are n constants then
E(SaX) - Sa E(X) proof or the theorem is mathematical induction.

Let us assume the theorem to be true for n = m

i.e. If Y = a1 X1 + a2X2 + ...............+ am Xm then we have

E(Y) = a1 E(X1) + a2 E(X2) .............+ am E(Xm)

Then E(a1X1 + a2X2 + ..................amXm+ am+1 Xm+1)

= E(Y+am+1Xm+1)

= E(Y) + am+1 E(Xm+1)

= a1E(X1) + a2E(X2) + ... + amE(XM) + am+1 E(Xm+1)

i.e. the theorem is true for n = m + 1

We know that the result is true for n = 1, n = 2

Hence it is true for all positive integral values of n

186

9.5.1 Covariance

If X and Y and two random variables, then covariance between them is defined as

Cov (X,Y) = E[(X-E(X)) (Y-E(Y))]

= E[XY - XE(Y) - YE(X) + E(X) E(Y)]

= E(XY) - E(X) E(Y) - E(Y) E(X) + E(X) E(Y)

= E(XY) - E(X) E(Y)

If X and Y are independent random variables then

E(XY) = E(X) E(Y). Hence Cov (XY) = 0.

Note: 1 Cov (aX, bY) = E[(aX-a x] (bY-by))

= E[a(X-x)b(Y-y]

= ab E[(X-X) (Y-Y)]

= ab Cov (X,Y)

2. Cov (X+a, Y+b) = Cov (X,Y)

3. Cov

9.5.2 Variance of Linear Combination of Random Variable

Theorem 6

If X1, X2,.................... Xn are n random variables then

Var
 187

Proof

Let U = a1X1 + a2 X2+..............+ a2 Xn

E(U) = a1E(X1) + a2 E(X2) + .... an E(Xn)

U - E(U) = a1 [X1 - E(X1)] + a2 [X2 - E(X2)] +.................... an [Xn - E(x2)]

Squaring and taking expectation on both side we get

E[U-E(U)]2 = + a12 E[X1 - E(x2)]2 + a22 E(X2-E(X2)]2

+......+ an2 E[Xn - E(Xn)]2

+2 a1aj E[{Xi - E(X1)} {X1 - E(X1)}}

i.e. Var U = a12 Var (X1) + a22 Var (X2) + ... + an2 Var (Xn)

+ aiaj Cov (Xi, Yj)

i.e Var aixj = a12 Var (X1) + 2 aiaj Cov (Xi, Yj)

Note

1. If a1 = 1, i = 1, 2, ....n

Var (X1 + X2 + ......... + Xn) = Var (X1) + Var (X2) + ... + Var (Xn) + 2, Cov (Xi Yj)

2 If X1, X2 ... Xn are pair wise independent random variable Then Cov (Xi, Xj) = 0 (i + j).
Hence we get Var (a1Xj + anXn) = a12 Var (X1) + ... + an2 Var (Xn)

In particular Var (X1 + ... +Xn) = Var (X1) + ........ + Var (Xn)

3. If a1 = a2 = 1 and a3 = a4 ... = an = 0 then we get

Var (X1 + X2) = Var (X1) + Var (X1) + Var (X2) + 2 Cov (X1, X2)
188

4. If a1 = 1, a2 = -1 and a3 = a4 ... = an = 0

Var (X1 - X2) = Var X1 + Var X2 - 2 Cov (x1, X2)

Thus Var (X + X2) = Var (Xt) + Var (X2) ± 2 Cov (X1, X2)

If X1 and X2 are independent

Var (X1 + X2) = Var (X1) + Var (X2)

9.6 Expectation of a Continuous Random Variable

X is a continuous random variable with p.d.f f(x) then

E(X) = xfx (x)dx = xdF(x) Provided the integral is absolutely convergent.

Theorem 7

If X and Y are continuous random variable

Then (1) E(X+Y) = E(X) + E(Y)

(2) E(XY) = E(X) E(Y) if X and Y are independent,

Proof

By definition E(X+Y)
 189

E(XY)

Problem 3

On the average, how many times must a die be thrown until one get a 4.?

Solution

Let X denote the number of throws of the die till he get 4. Then X take the value 1, 2, 3....
with the respectively profitability 1/6, 5/6 (1/6), (5/6)2 (1/6)....

~E(x) = (1/6) + 2(5/6) (1/6)+3(5/6)2 1/6 + 4 (5/6)3 1/6 +

= 1/6[1+2(5/6) + 3(5/6)2 + 4(5/6)3 + ....................]

= 1/6[1+2x + 3x2 + ....] where x = 5/6

= 1/6 [1-x]2

= 1/6[1-5/6]2 = 1/6 (1/6)2 = 1/6 X 62 = 6.

9.7 Moment Generating Function

The moment generating function m.g.f of a random variable X about origin have the
probability function f(x) is given by

Mx(t) = E(etx) = etx f(x) dx for continuous probability distribution.

= x etx p(x) for discrete probability distribution.

The integration or summation is extended to the entire range x, t being real parameter
and it is assumed that the right hand side give above is absolutely convergent.
190

Thus Mx (t) = E(etx)= E(1+tX+t2 X2/2!+...+tfXr/r!+...)

= [1+tX+(t2/2!) E(X2)+...+(tr/r!)E(Xr) +

= 1+tµ'1 + t2µ12/2! + .... tfµ'r/r! + ...(A)

where µ'r = xf xf p(x) for discrete distribution

= x' f(x) dx for continues distribution is the rth moment of X about origin. So the coefficient
of tr/r! in Mx (t) give µ'r. Since Mx (t) generates moments of different order it is called as moment
generating function.

Differenting (A) w.r.t. and putting t = 0

Theorems on Moment Generating Function

Theorem 8

Mcx (t) = Mx (ct)

Proof

By definition Mcx(t) = E(etx)

Mx(ct) = E(etx)

~L.H.S= R.H.S

Theorem 9

The moment generating function of the sum of a number of independent random variable
is equal to the product of their respective moment generating functions.

If X1, X2, ........... Xn are n independent random variables then the moment generating
function of their sum X1 + X2+...... + Xn is given by
 191

Mx1 + x2+.....+xn (t) = Mx1(t).......Mxn(t)

Proof by definition

Mx1 + x2+..........+ xn(t) = E[e 1(x1 + x2+.....xn)]

= E [etx1 etx2...etxn]

= E (etx1) E(etx2).... E(etxn)

(Since X1, X2......... Xn are independent)

= Mx1 (t) Mx2 (t) ..... Mxn (t)

Theorem 10

Effect of change of origin and scale on m.g.f.

Let us transform X into a new variable U by changing the origin and scale in X as

U = x-a/h where a,b are constants.

[Link] of U about origin = Mij(t) = E(etu) = E[et(x-a)/h]

= E [eth/h e-a/th]

= e-ath E(etx/h)

= e-at/h Mx(t/h)

where Mx (t) is the m.g.f of X about origin.

Taking a = E(X) = µ (say) and h = x =  (say) then

X-E(X) X-u
U= = = z say
x 


is known as standard variate.

The m.g.f. of a standard variate Z is given by Mz (t) = e-tµ/-Mx(t/)

192

E(Z) = 0 and Var Z = 1. That is the mean and finance of a standard variate are 0 and 1.

Problem 4

A random variable X has the density function

f(x) = X+1/2 for-1 =x=1

=0 else where

Find the moment generating function of x.

Solution
 193

Cumulants

The n cumulant to defined as the coefficient of tr/r! in logo Mx (t)

9.8 Characteristic Function

The characteristic function is defined as

x (t) = E(eitx) = p(x) for discrete probability distribution

= eitx f(x) dx for continuous probability distribution.

For a distribution m.g.f. Mx(t) may or may not exit but characteristic function Øx(t) always
exist. This is advantage of characteristic function over the moment generating function.

Check Your Progress

Fill in the blanks

1. If x is a continuous random variable with p.d.f f(x) then the Ex ___________________

2. Mx(t)____________________for discrete probability distribution.

3. The characteristic function os defined as ___________________

9.9 Summary
In this lesson, we have learnt the conditional expectation, moment generating function
and characteristic function.

9.10 Check Your Answers

1. vxf(x).dx

2. etx P(x)

3.  x(t)
194

9.11 Model Questions

1. A man draws 2 ball from a bag containing 3 white and 5 blacks balls. If he receives Rs.70
for every white ball and Rs.7 for every black ball he draw. Find his expectation. (Ans.
Rs.61.25).

2. On the average how many times must a die be thrown until one get a '6'. (Ans. 6)

3. A coin is tossed until a 'head' appears. Find the expected number of tosses. (Ans. 2)

4. Two unbiased dice are thrown. Find the expected value of this sum of numbers of point
on them.

5. Find the mathematical expectation and the variance of the sum of points on n dice. (Ans.
7n/2, 35n/12).
 195

LESSON - 10
DISCRETE DISTRIBUTIONS
Structure

10.1 Introduction

10.2 Learning Objectives

10.3 Bernouli Distribution

10.4 Binomial Distribution

10.5 Poisson Distribution

10.6 Summary

10.7 Check Your Answers

10.8 Model Questions

10.1 Introduction
The statistical data collected are presented in the form of a frequency distribution. These
distributions are based on actual data. But there are certain distributions which are not based on
actual data or experiments, but they are derived mathematically or theoretically on the basis of
certain assumptions. Hence these distribution are called theoretical distributions. This broad
classification under which these are studied discrete and continuous theoretical distribution.
This lesson will deal with only two Discrete Distributions 1. Binomial Distribution, 2. Poisson
Distribution.

10.2 Learning Objectives

After studying this lesson, you should be able:

• to know about the form of Binomial and Poisson distribution; Parameters, moments,
generating function, characteristic function of the Binomial and Poisson distribution

• to do problems connected with Binomial and poisson distribution.

• to calculate the expected frequencies of a Binomial and poisson variant.

196

10.3 Bernoulli Distribution

A random variable X which takes two values 0 and 1 with probability q and p respectively.
i.e., P(X=1) = p, P(X=0)=q, q=1-p is called a Bernoulli variable and is said to have a Bernoulli
distribution.

10.4 Binomial Distribution

This distribution was discovered by James Bernoulli. Let a random experiment be
performed repeatedly and let the occurrence of an event in any trial be called a success and let
its non occurrence is called a failure. Consider a series of n (n being finite) independent Bernoullian
trials, in which the probability p of success in any trial is constant for each trial. Then q=1-p is the
probability of failure in any trial.

The probability of x success and (n-x) failures in n independent trails in a specified order
say SFSFFS...FS, where S represent success and F denotes failure, is given by the product
theorem of probability as

p(SFSFFS...FS) = p(S)p(F)p(S)p(F)p(F)...p(S)

= pq pqq...p

= (p Xp X. x factors)

= x (qXqX - (n-x) factors

= pxqn - x

But the x success in n trails can be selected from the n trails in nCx ways and the
probability for each of these ways = pqnx. Hence the probability of x success in n trails, in any
order what so ever is given by the addition theorem of probability by the expression nCxpx qn-x.

The probability distribution so obtained is called the Binomial distribution. Giving values
for n as 0, 1, 2..... n, the respective probabilities are qn, nC1pqn-1, nC2p2qn-2...pn which are the
successive terms of the Binomial expansion (q+p)n.

10.4.1 Probability density function

A random variable X is said to follow the Binomial distribution, if its probability density
function is given by
 197

n
P(X=x) = Cxpxqn-x x = 0, 1...n

= O elsewhere

The two constants n and p are called the parameters of the distribution. Since X can
take only positive Integral values Binomial distribution is a discrete distribution. Any variable
which follow the Binomial distribution is called a Binomial variable. If X is a binomial variable with
parameters n and p. We shall write it as X~ B (n-p).

If the n trails constitute an experiment and if the experiment is repeated N times then the
frequency function of the binomial distribution is given by p (X=x) = N nCx pxqn-x and the successive
frequencies for x = 0, 1.... are given by the successive terms of the expansion N (q+p)n.

Binomial distribution has wide applicability and give raise to many other probability
distributions.

Problem 1

A machine manufacturing screws is known to produce 5% defectives. In a random sample

of 15 screws what is the probability that there are (1) exactly three defective and (2) not more
than three defectives.

Solution

Let p denote the probability for getting a defective screws

p(x) = 5/100.. q = 1-p = 95/100

Hence the probability for getting 3 defective screws out of 15 screws is 15C3 (5/100)3 (95/
100)12

15
P(X=0) = C0 (5/100)0 (95/100)15

15
P(X=1) = C1 (5/100)1 (95/100)14

15
p(X=2) = C2 (5/100)2 (95/100)13

15
p(X=3) = C3 (5/100)3 (95/100)12

Hence p(X=0) or p(X=1) or p(X=2) or p(X=3)

198

= (95/100)15+15(5/100) (95/100)14

+105 (5/100)2 (95/100)13 + 445 x (5/100)3 (95/100)12

n(n-1)...r factors
nc, =
1.2.3....r factors

= (95/100)12 [(95/100)3 + 5 x (5/100) (95/100)2

+ 105 x (5/100)2 x (95/100)] + 455 x (5/100)3

Hence the probability of getting not more than 3 defective is 1-(95/100) [(95/100)3 x
(5/100) (95/100)2 + 105 (5/100)2 (95/100) + 455 (5/100)3]

Problem 2

Taking 100 sets of 10 tosses of a perfect coin. In how many sets do you expect to get
atleast 7 heads.

Solution

In the case of perfect coin p = q = 1/2.

Since n = 10 and N = 100, the frequency corresponds to

7 head is 100 [10C1 (1/2)7 (1/2)3]

The frequency corresponding to 8 heads is

 199

The frequency for 9 success is

The frequency for 10 success is

100 [10C10, (1/2)10 (1/2)0]

= 100 x 1 x (1/210)

= 100 / 1024

= 25/256

Hence the frequency for getting atleast 7 head is

= 375/32 + 1125 /256 + 125/128/ + 25/256

= 11.72 + 4.40 + 0.98 + 0.10

= 17.2

Problem 3

A purchaser will accept of lot of 120 articles if a sample of 5 articles picked up at random
from the lot contains not more than one defective. What is the probability that he will accept the
lot if it contain 15 defective articles.?

Solution

Probability that a randomly chosen articles is defective = 15/120 = 0.125. Choosing 5

articles randomly is the same as repeating the trial 5 times in which the probability of defective
and good are 15/120 and 105/120. Hence the probability of getting no defective in the five trails
is 5C0(15/120)0 (105/120)5 = 105/1285

Probability of getting one defective = 5C1 (15/20)1 (105/120)4

Hence the required probability = [105/128]5 + (75/20) {105/120}4

200

Problem 4

The probability of a man hitting a target is 1/3. If he fires 8 times. (1) What is the probability
of him hitting the target atleast twice? (II) How rnany times must he fire so that the probability of
his hitting the target atleast once is greater than 1/2?

Solution

Let p be the probability of him hitting the target, then p = 1/3 and q = 1-(1/3) = 2/3

(1) Probability of hitting the target at least twice in 8 trails is

P(x > 2) = 1-p (X < 2)

= 1-p (X=0) + p(x=1)

= 1- 8C0 (1/3) (2/3)8 - 8C, (1/3)1 (2/3)7

= 1-1/3 x (2/3)8-8 x 1/3 x (2/3)7

= 1-1/3 x 256/661-8/3 x 128 / 2187

= 1-0.013-0.156

= 1-0.169

= 0.831

(II) Let n be the required number of time he must fire to fit the target atleast once, ther the
probability for it 1-qn > 1/2

i.e., 1-(2/3)n > 1/2

i.e., (2/3)n < 1/2

~n=2

10.4.2 Moments

µ'1 = E(X) =
 201

Hence the mean of Binomial distribution = np

202

Similarly

X4 = x(x-1) (x-2) (x-3) + 6x (x-1) (x-2) + 7x (x-1) + x

= E(x4) = x4 nCxpxqn-x

= x(x-1) (x-2) (x-3) nCx (pxqn-x +

= 6x(x-1) (x-2) nCx pxqn-x +

= 7 x(x-1) nCx pxqn-x + x nCx pxqn-x

On simplification, we get

   n(n-1) (n-2) (n-3) p4 + 6n(n-1) (n-2) p3 + 7n(n-1) p2 + np

10.4.3 Central Moments

µ2 = µ12 - (µ11)2 = n2p2 - np2 + np - n2p2

= np(1-p) = ?pq

µ3 = µ'3 -3µ2' µ'1 +2µ,3

= [n(n-1) (n-2) p3 + 3n (n-1) p2 + np]

-3 [n(n-1) p2 + np] np + 2 (np)3

= np [-3np2 + 3np + 2p2-3p+ 1-3npq]

= np [3np (1-p) + 2p2-3p+ 1-3npq]

= np (2p2 - 3p + 1)

= np x (2p2 - 2p + q)

= np q x (q-p)

µ4 = µ4 - 4µ3 µ1 + 6µ2 µ12

= npq [1+3 (n-2) pq]

 203

Hence

10.4.4 Recurrence Formula for Central Moments of a Binomial Distribution

The recurrence formula enable us to calculate the various central moments easily.

We have µr = E(x-µ)

= (x-np)r nCx Px qn-x

dµr/dp = r(x-np)r-1 (-n) nCx pxqn-x + (x-np)r nCx xpx-1qn-x

+ (x-np)r nCx px (n-x) qn-x-1 (-1)

= - r n µr-1,+ (x-np)r (nCx) pxqn-x [x/p) + (n-x)/q]

= -r n µr-1 +(1/pq) µr+1

~µr+1 =

Knowing that µ0 = 1, µ1 = 0 we can easily calculate all higher order moments given
Binomial Distribution.
204

For example

µ2 = pq (nµ0 + (d/dp) µ1) = npq

µ3 = pq (2nµ1 + (d/dp) µ2) = p q (n p (-1) + nq)

n p q (q-r) etc.

10.4.5 Moment Generating Function

Let X be a random variable following Binomial distribution

M (t) = E(etx) = etx nCx px qn-x

= (p et)x qn-x nCx = (q + pet)n

M.G.F about mean for binomial distribution

E(e t(x-np) = E(et e-t n p) = e-t n p E(etx) = e-t n p Mx (t)

= e-tnp (q+pet)n

= (qe-pt + petq)n
 205

µ2 = Co-efficient t2/2! in M.G.F = npq

µ2 = Co-efficient of t3/3! = ?pq (q-p)

µ2 = Co-efficient of t4/4! = npq (1-3pq) + 3n(n-1) p2q2

= npq (1-3pq) + 3n2p2q2 - 3np2q2

= 3n2p2q2 + npq (1-6 pq)

10.4.6 Characteristic Function of Binomial Distribution mobod

Ø x(t) = E(eitx) = p(x)

= x nCxpx qn-x

= (nCx) (peit) xqn-x

= (q + pe it)x

10.4.7 Recurrence Formula for Probability of Binomial Distribution

206

This formula given as easy method to calculate probabilities of a Binomial Distribution.

Knowing p(0) = qn where q = 1-p and p = x/n we can find the remaining probabilities.

10.4.8 Mode of the Binomial Distribution

Mode is the value for which p (x) is maximum

Case i When (n+1) p is not an Integral

Let n be the Integral Part of (n+1) p

We have, by putting x = 1, 2 ..... n

and

~ P(0) P(1) < P(2) ... <P (m-1) <P(m) > P(m+1) > P(m+2). > P(n) and so on. Thus in
this case there exist a unique model value and it is m the Integral part of (n+1).
 207

Case 2; when (n+1) p is an Integer.

Let (n+1)p = m. Proceeding as in case (1) we have p(0) <p(1) .... <p (m-1) = p(m)>
p(m+1) > p(m+2) ... > p(n). Thus in this case the distribution is bimodal, modal values are
m and m-1.

Problem 5

Assuming that 20% of the population of a city are literate and assuming that 100
investigators each take 10 individuals to see whether they are literate how many investigation
would you expect to report that three people or less were literate?

Solution

Let x denote the number of literate

Then we require p(x = 0) + p(x=1) + p(X=2) + p(X=3)

Since p = 20/100 = 1/5

q= 1-p = 4/5

10
P(X=0) = C0 (1/5)0 (4/4)0 = 0.107

10
p(x=1) = C1 (1/5)1 (4/5)9 = 0.268

10
p(x=2) = C2 (1/5)2 (4/5)8 = 0.302

10
P(X=3) = C3 (1/5)3 (4/5)7 = 0.201
0.878

The number of investigators to report for 3 or less

literatures = 0.878 x 100 = 87.8 or 88

Problem 6

The mean and variance of a Binomial Distribution are 4 and 4/3 respectively. Find p
(x = 1)
208

Solution

Mean = np = 4

Variance = npq = 4/3

4/3
q = = 1/3
4

p = 2/3

Since np = 4, We have n x 2/3 = 4

n = 6

The required Binomial distribution is (1/3 + 2/3)6

P (X = 1) = 1-p (x=0)

= 1-qn

= 1 - (1/3)6

= 1- (1/729)

= 1 -0.00137

= 0.99863

Problem 7

A discrete variable takes the values 0, 1, 2....n with frequencies proportional to 1, nC1,
n
C2, ... nCn respectively. Show that the mean is n/2 and variance is n/4

Solution

Let k be the constant of proportionality, then the total frequency = K(1 + nC1 + nC2+
..+nCn)

= K(C0 + C1 + C2 + ... + Cn)

= K.2n (since the sum of binomial coefficient = 2n)

Hence the probability for 0, 1, 2...n values of the variate are respectively.
 209

~ The binomial distribution is (1/2 + 1/2)"

Here p = q = 1/2

Hence mean = np = n/2

Variance = npq = n/4

Problem 8

Five dice were thrown 96 times. The number of times 4, 5 or 6 actually thrown in the
experiment is given below. Calculate the standard deviation and the quartile deviation of the
observed frequnces. What is the theoretical value of the standard deviation. Calculate the
expected frequencies.

Number of dice showing

4, 5 or 6 0 1 2 3 4 5

Observed Frequency 1 10 24 35 18 8

Solution

X f fx fx2 end x Cum. Freq

0 1 0 0 0.5 1

1 10 10 10 1.5 11 Q1

2 24 48 96 2.5 35

3 35 105 315 3.5 70 Q3

-4 18 72 288 4.5 88

5 8 40 200 5.5 96

96 275 909
210
 211

Calculation of Expected Frequencies

Check Your Progress - I

Fill in the blanks

a. Binomial distribution is one of the _____________distribution.

b. Binomial distribution depends upon the values of _______________and ___________

c. The standard deviation of______________is npq

10.5 Derivation of Poisson Distribution

Poisson distribution is the limit form of binomial distribution when n   and p  0 so
that np is fixed quantity = m

In the binomial distribution, the probability for r success is given by P(r) = nCr Pr qn-r We
write this as

By making n, we have

212

Using Striling's formula for n!., we have

since er. e-r = 1, r being finite

The probability for 0, 1, 2...... success are

e-m, me-m m2 e-m ,..respectively.

This probability distribution is called. Poisson distribution.

10.5.1 Probability density function

A random variable X is said to follow Poisson distribution if it assumes only non- negative
values and its probability mass function is given by

= 0 otherwise

m is called the parameter of the distribution and m > 0.

The sum of the probability =

The distribution has only one parameter m. Since p? 0, this distribution is also known as
distribution of rare events.

1. The number if deaths in a district in one year by a rare disease.

2. The number of defective screws per box of 100 screws.

 213

3. The number of typographical errors per page in type material.

4. The number of death in a big city due to road accident in a year.

These are examples of a Poisson variate.

Another Derivation of Poisson Distribution

Problem 9

If X is a Poisson variate such that P(X=1) = P(X=2), find P(X=4).

Solution

Since P(X=1) = P(X=2), we have,

214

10.5.2 Moments of Poisson Distribution

Hence the mean of the Poisson distribution is m.

 215

Hence the first four central moments are

µ2 = µ'2- (µ'1)2 = (m2 + m) - m2 = m

Thus the mean and variance of the Poisson distribution are equal, equal to m.

µ2 = µ'3- 3µ'2µ'1 + 2µ'13

= (m3 + 3m2 + m) - 3m(m2 + m) + 2m3

= m
216

µ4 = µ14- 4µ13µ'1 + 6µ'2µ'12 - 3µ'14

= (m4 + 6m3 + 7m2 + m) - 4m(m3 + 3m2 + m) + 6m2 (m2 + m) - 3m

= 3m2 + m

Co-efficient of skewness

Hence the Poisson distribution is always a skewed distribution.

Putting m , we get , = 0 and 2 = 3.

Hence Poisson distribution tends to normal distribution.

10.5.3 Recurrence Relation for the Moments of the Poisson Distribution

By definition

Differentiating with respect to m, we get,

 217

Putting r = 1, 2 and 3 successively, we get

10.5.4 Moment Generating Function of the Poisson Distribution

10.5.5 The Characteristic Function of the Poisson Distribution

218

Additive Property of Independent Poisson Variate

Theorem

Sum of two independent Poisson variates is also a Poisson variate. If X1 = (i = 1, 2..n)

are independent Poisson variates with parameters m1, m2,..... mn respectively, then

X is also a Poisson variate with parameter ma

Proof

Mxi(t) = emi (et-t) i = 1, 2....

Mx1,+x2+...xn (t) = Mx1(t).Mx2 (t).................... Mxn (t)

(Since x1; i = 1, 2...n are independent)

= em1 (et-1)
em2 (et-1).... emn(et-1)

=e (m1+m2+ m3..+mn) (e1-1)

This being the moment of generating function of a Poisson variate with parameter
m1+m 2+..+mn.

By uniqueness theorem of moment generating functions X, is also Poisson variate with

m1.

Note that the difference between two independent Poisson variates is not a poisson
variate.

Problem 10

If X and Y are independent Poisson variates such that P(X=1) = P(X=2) and P(Y=2)=
P(Y=3), find Var (X-2Y).

Solution

Let X and Y follow the Poisson distribution with parameters m and , then
 219

Using the given conditions, we get,

From p(x = 1) = p(x = 2)

we get i.e., m = 2

p(Y = 2) = p(Y = 3)

and since X and Y are independent, covariance term vanishes.

So, we get Var(X-2Y) = 2 + 4x3 = 14.

Problem 11

An auto-rickshaw firm has two auto-rickshaws, which it hires out day by day. The number
of demands for auto-rickshaw follows the Poisson distribution with mean 1.8. Calculate the
proportion of day on which

1. Neither autos are used

2. Some demand is refused.

220

Solution

Since the number of demands for and auto on any day is a Poisson variate with mean
1.8, the proportion of days on which there are x demands for an auto.

1. Proportion of days on which neither auto is used is given by

P(X = 0) = e-1.8

= 0.1652

2. Proportion of days on which some demand is refused is

P(X >2) = 1- P(X =2)

= 1 - [P(X=0) + P(X = 1) + P(X = 2)]

= 1-e-1.8

= 1 -0.1652[2.18 +1.62]

= 1 -0.6278

= 0.3722

Problem 12

A random variable following Poisson distribution takes the values 2 and 3 with equal
probability. Determine the probability that the value taken by the random variable is less than or
equal to 3.

Solution

Let m be the parameter of the Poisson distribution.

Given p(x = 2) = p(X = 3)

 221

Hence the probability for (X = 3) i.e. P(X =3)

= p(x = 0) + P(X = 1) + P(X = 2) + P(X = 3)

Problem 13

In a certain factory turning out razor blades, there is small chance 1/500 for any blade to
be defective. The blades are supplied in packets of 10. Use the Poisson distribution to calculate
the approximate number of packets containing no defective, one defective and two defective
blades respectively in a consignment of 10000 packets.

Solution

Let X be the number of defectie blades.

Given p = 1/500 and n = 10

Thus m=np= 1/500 × 10 = 1/50 = 0.02

e-0.2(0.02)x
So p(x = x) =
x!

e-0.2(0.02)°
P(X = 0) = =e-0.2 = 0.9802
0!

Hence the number of packets containing no defective blades

= 10000 x 0.9802 = 9802

222

P(X = 1) = e-0.2(0.02)1
1!

= 0.9802 x 0.02

= 0.019604

Hence the number of packets containing one defective blades

= 0.00019604 x 10000 = 196.

10.5.6 Recurrence Formula for the Probabilities of the Poisson Distribution

For the Poisson Distribution

Knowing only the probability P(0)= e-m, which is estimated from the data, using the
recurrence relation we can find the probability for x = 1, x = 2...

Problem 14

A typist commit the following number of mistakes per page in typing 100 pages.

Number of mistakes per page 0 1 2 3 4 5

Number of pages 48 30 14 4 3 1
 223

Solution

X f fx

0 48 0

1 30 30

2 14 28

3 4 12

4 3 12

5 1 5

Total 100 87

The required Poisson distribution

P(X = x) = e-0.87 (0.87)x Expected Frequencies

X

P(X = 0) = e-0.7 = 0.419 42

(0.87)
P(X = 1) = x 0.4190 = 0.3645 36
1

(0.87)
P(X = 2) = x 0.3645= 0.1586 16
2
(0.87)
P(X = 3) = x 0.1586= 0.0460 5
3
(0.87)
P(X = 4) = x x 0.046 = 0.01000 1
4
(0.87)
P(X = 5) = = x 0.01 0.00174 0
5
224

10.5.7 Mode of Poisson Distribution

The mode is the value of r for which is greater than the term that precede and

suceed it

This gives r < m and r+1 > mi.e., m-1 < r < 1. If m is not an integer, then the mode is the
integral value between m-1 and m. If m is an integer the probability for the Poisson variate to
take the value m-1 or m is the same. Hence, both can be taken as the mode. In this case the
distribution is bimodal.

Check Your Progress - II

Fill in the blanks

a. __________________is called the parameter of the Poisson distribution.

b. Poisson distribution is also known as distribution of _______________

c. The difference between the two independent poisson variate is ______________

10.6 Summary
The discrete distribution have different types. In this lesson we have learnt Bernoulli,
Binomial and Poisson distributions.

10.7 Check Your Answers

I. a. Discrete type b. n, p, q c. Binomial distribution

II. a. Mean b. rare events c. Not Poisson variate

10.8 Model Questions

1. Five men in a company of 20 are graduates. If 3 men are picked out of the 20 at random
(1) what is the probability that there is atleast one graduate? (2) what is the probability
that they are all graduate?
 225

2. Find the binomial distribution where mean is 9 and variance is 2.25

3. 15% of the bolts produced by a machine are defective. Find the produced by a machine
are defective. Find the probability distribution of defectives in a sample of 3 bolts chosen
as random.

4. A machine manufacturing screws is known to produce 5% defectives. In a random sample

of 15 screws. What is the probability that there are (i) exactly three defectives (ii) not
more than three defectives?

5. In a binomial distribution consisting of 5 tirals independent probability of 1 and 2 successes

are 0.4096 and 0.2048 respectively. Find the parameter p of the distribution.

6. With usual notation find p for a binomial random variable X if n = 6 and if 9 p (X=4) = P
(X=2).

7. A set of 8 symmetrical coins were tosed 256 times and the frequencies are as follows:

Number of heads 0 1 2 3 4 5 6 7 8

Frequence 2 6 24 63 64 50 36 10 1

Find the expected frequencies.

8. A hospital switch board receives an average of 4 emergency calls in a 10 minute interval.

What is the probability that

i) There are at the most 2 emergency calls in a 10 minute interval

ii) There are exactly 3 emergency calls in a 10 minute interval.

9. It is 1 in 1000 that a twin birth takes place. In a town on one day 100 births occur. Find the
probability that one or more twins are born. a) using binomial law b) using Poisson law.

10. A manufacturer of cotter pins knows that 5% of his product is defective. If he sells cotter
pins in boxes of 100 and guaranties that not more than 10 pins will be defective, what is
the approximate probability that a box will fail to meet the guaranteed quality?

11. Fit a Poisson distribution for the following data which gives the number of yeast cells per
square for 400 squares.

[Link] cells per square 0 1 2 3 4 5 6 7 8 9 10 Total

No. of squres 103 143 98 42 8 4 2 0 0 0 0 400

226

LESSON - 11
CONTINUOUS DISTRIBUTIONS
Structure

11.1 Introduction

11.2 Learning Objectives

11.3 Normal Distribution

11.4 Expontial Distribution

11.5 Summary

11.6 Check Your Answers

11.7 Model Questions

11.1 Introduction
Continuous distribution is one of the theoretical distribution. Here we shall discuss the
Normal and Exponential distribution.

11.2 Learning Objectives

After studying this lesson you should be able to

• know about Probability Density Function of a Normal variate

• understand the importance of Normal & Exponential distribution Normal Probability Curve,
moments and moment generating function etc.

• to work out problems of Normal distribution

• calculate expected frequencies

11.3 Normal Distribution

The Normal Distribution was first discovered by English Mathematician De-Moivre, who
obtained this continuous distribution as a limiting case of the Binomial distribution and applied
 227

to the problem arising the game of chance. Normal distribution is the most widely used probability
model for continuous random variables.

Definition

A random variable X is said to have a normal distribution with parameterand 02 (called

mean and variance respectively) if its density function is given by the probability law.

A random variable X with mean ì and variance o2 and which follows the above normal
law is represented by X-N(, 02).

is a standard normal variate with E(Z) = 0 and Var(Z) = 1 and we write Z~N(0,1).

The probability density function of standard normal variate Z is given by

The Fig 11.1 of f(x) is a bell-shaped curve. The top of the ball is directly above the mean ì.

Median Mode Mean

Fig. 11.1 Normal Curve

228

11.3.1 Normal Distribution as Limit Form of Binomial Distribution

Normal distribution is another limiting form of binomial distribution under the condition.

i) n, the number of trails is indefinitely large i.e., n’! “

ii) neither p nor q is very small.

The probability function of distribution with parameters n and p is given by

Consider the standard binomial variate

Thus in the limit as nZ takes values from - to . Hence the distribution of X is a
continuous distribution in (-, ). To get the limiting from of binomial distribution, we use Stirling’s
approximation to r! for large r.

We have in the limits as n  and Consequently x 

 229

Hence the probability of the distribution of Z, in the limit is given from (3) by
230

Substituting in (4), we get

Hence the probability function of Z

..

This is the probability density function of the normal distribution with mean 0 and
unit variance.

If X is a normal variate with mean  and

standard deviation  then Z = X -  / is a standard normal variate.

Jacobean of the transformation 1/Hence substituting in (5), the, probability density

function of a normal variate X with
 231

11.3.2 Characteristics of the Normal distribution and normal probability curve

The normal probability curve with mean ì and standard deviation ó is given by the equation.

and has the following properties

i) The curve is bell shaped and symmetrical about the line x = ì

ii) Mean, Median and Mode of the distribution coincide.

iii) As x increase numerically, f(x) decrease rapidly, the maximum probability occupying at
x =.

iv) 1 = 0; 2 = 3

v) All the odd order central moments vanishes i.e.,

2r+1 = 0 (r = 0, 1, 2........) and

2r = 1.3.5.... (2r-1)2r (r = 0, 1, 2.....)

vi) Linear combination of independent variates is also a normal variate.

232

vii) x axis is an asymptote to the curve

viii) The point of inflexion of the curve are at

ix) Mean deviation about Mean is

(approximately)

x) P(-x ) = 0.6826

P(-x ) = 0.9544

P(-x ) = 0.9973

xi) Quartile deviation for normal distribution = 2/3

11.3.3 Moments

Mean of Normal Distribution

Mean

put

Since the integral being odd function of z;

 233

Variance of Normal Distribution

Var X = E(x - )2

= (x - )2 f(x)dx
234

Mean Deviation

The mean deviation from the mean 

Moments about the mean 

(since the integrand is an odd function of z)

Thus all odd order moments about the mean vanish

 235

Integrating by parts

Applying the recurrence formula successively, we obtain

2n = (2n-1) (2n-3)............ 3. 12n.

In particular

2 = 2 2 = 34

11.3.4 Moment Generating Function About Origin

Moment generating function with respect to mean ì.

236

11.3.5 Recurrence Relation

Median and Mode of Normal Distribution

Let ‘a’ be the median of the normal distribution, then

 237

as easily seen by putting x = .

Hence the median of the normal distribution is equal to the mean.

ii) We have f(x) =

Taking logarithm on both sides,

Differentating both sides with respect to x

Hence f’(x) = 0 gives x = ì

and [f’(x)]xu = - 1/2 [f(x)]x = u

Hence the mean ì is the mode of the normal distribution.

A linear culmination of independent normal variates is also normal variate

Let X1 = 1, 2, 3...n, be n independent normal variates with mean , and variance 12
respectively. Then,

The moment generating function of their linear combination where a,. a,...... a

are constants given by

238

which is the moment generating function of a normal variate with au

Hence by the uniqueness theorem

Thus the sum as well as the difference of two normal variates is a normal variate.

Taking a1 = a2........an = 1, the

 239

i.e. the sum of independent normal variate is also a normal variate. The establishes the
additive property of normal distribution.

Letx, ; i = 1,2..n be n independent normal variates with mean ì and variance o2 and if we
take a1= a2 = .. an = 1/n then,

Let X1 ; i = 1,2..n be n independent normal variates with mean ì and variance ó2 and if
we take a1= a2 = .. an = 1/n then,

Thus, if X1, ; i = 1, 2 ..n are n identically distributed independent normal variates with and
variance o2, then their mean

Point of Inflexion of Normal Curve

At the point of inflexion f’(x) = 0 and “‘(x) = 0.

For normal curve, we have

At these points f’” (0) 0. Hence the point of inflexion are at x = and the ordinate at
those points is
240

11.3.6 Area Property

If X~N , then the probability that the random variable X will lie between x =  and x-
x, given by

is the probability function of standard normal variate.

The definite integral

is known as the normal probability integrand. It gives the area under standard normal
curve between the ordinate at Z = 0 and Z = Z, these areas are calculated for different values of
Z, and given in the form of a table.

In particular the probability that a random variable X lies in the interval is given
by
 241

Fig. 11.3 Area Property

Similarly

P(-2 < X < + 2) = 0.9544

P(-3 < x < + 3) = 0.9973

Hence the probability for X to lie outside the interval  - 3ó,  + 3 is 1-0.9973 = 0.0027.
Since the are under normal probability curve is unity. The area of the normal probability curve to
the left of Z= 0 is equal to the area of the normal probability curve to the right of Z = 0.
Each = 0.5.
242

Problem 1

Assume that the mean height of soliders to be 172 cms and the variance (27 cm)?. How
many soldiers in a regiment of 1000 can be expected to be over 182 cms.

Solution

We require P(X > 182) given x = 172 2 = 27

Problem 2

The standard deviation of a certain group of 100 high school grades was 11% and the
mean grade 78%. Assuming the distribution to be normal, find,
 243

1. How many grades were above 90%?

2. What was the highest grade of the lowest 10?

3. What was the inter-quartile range?

4. Within what limits did the middle 90% lie?

Solution

Here  = 78  = 11

1. We require

Number of students getting grades above 90% is 0.1379 x 100 = 137.9=138

2. Let x be the highest grade of the lowest 10%, then, Z =

The area to the left of Z is 10/100 = 0.1

244

3. Let Q1 be the first quartile. Then the area between

Referring to the table of area Z1 = -0.672 i.e.,

 245

Let x1 and x2 be the limits with which middle 90% lie

Then the area of the normal probability curve between Z1, and 0 is 0.45. Referring to the
table of areas
246

Problem 3

The following table gives the frequency of occurence of a variate X between certain

Variate (X) Less than 40 40 or more but less than 50 50 and more Total

Frequency 30 33 37 100

The distribution is exactly normal. Find the averate standard deviation of x and hence
the frequency between x = 30 and x = 40.

Solution

Let x and be the mean and standard deviation of the distribution.

Then the area of the standard normal probability curve to the left of Z1 = 30/100 = 0.3

~Hence the area between Z = 0 and Z = Z, = 0.2

~Z1 = 0.522. (from table of area)

Let Z2 = 50-x / 

Then the area of the standard normal probability curve to the right of Z2 = -0.37

Hence the area between Z = 0 and Z = Z2 = 0.13

~ Z2 = 0.332 (from table of areas)

 247

Subtracting 10 = 0.854

The area between Z = 0 and Z = -1.39 = 0.4177

The area between Z = 0 and Z = 0.53 = 0.2019

The are between Z=-1.39 and Z = 0.53 0.2158 x

Number of deservations between, 30 and 40 is 0.2158 x 100 = 21.58 i.e. 22.

Problem 4

In a distribution which is exactly normal 31% of the items are under 45 and 8% over 64.
What are the mean and standard deviation of the distribution.
248

Let X and a be the mean and standard deviation of the distribution.

Let Z1 = Then the area to the left of Z, = 0.31

Hence the area between Z = 0 and Z = Z, = 0.19

Hence the area between Z = 0 and Z = Z, is 0.5-0.08 = 0.42

 249

Problem 5

A minimum height is to be prescribed for eligibility to Government services such that

60% of the young men will have a fair chance of coming up to that standard. The height of
young men are normally distributed with mean 60.6" and standard deviation = 2.55". Determine
the minimum specification.

Solution

Let x be the minimum height.

Then Z1 =

The area to the right of Z1 is 0.60. Hence the area between Z = 0 and Z = Z1 is 0.1.
Referring the table of area it is -0.251.

11.3.7 Importance of the Normal Curve

The normal curve has an important place in Statistical theory for the following reasons.

1. Errors of observation are normally distributed.

2. Biometrical data are found to follow the normal law more or less approxmiately.

3. Binomial curves resemble the normal curve for values of n greater than 30.

4. It is only with reference to the normal curve that the term skewness and kurtosis are
defined.

5. The normal curve is found to be a very close approximation to bell-shaped curves.

6. Most of the distributions that occur in nature are found to be more or less normal if
not exactly normal.

7. The sampling distribution of many of the population statistic are found to be normal for
large samples.
250

Problem 6

The table below show the frequency distribution of diameter of 67 trees standing on a
plot of land.

Diameter in Inches 7 8 9 10 11 12 13 14 15

Frequency 1 6 7 11 20 10 6 5 1

Fit a normal curve to the data and calculate the expected frequencies.
 251

Check Your Progress - I

Fill in the blanks

a. Normal curve is ______________shaped curves.

b. Mean, Median and Mode of the distribution_ _______________

11.4 Exponential Distribution

A continuous random variable x assuming non-negative values is said to have exponential
distribution with parameter 0 > 0, if its p d f is given by
252

Any variate possessing the pdf (1) is expressed as X ~ Exp (0).

cumulative distribution function F (x) is given by

11.4.1 Moments
 253

It follows the for Exp (0), Mean <, = >, variance according as >, < 1. This is a remarkable
feature of this distribution.

11.4.2 Moment Generating Function

Expanding the binomial expression involved

11.4.3 Characteristic Function

We may proceed as above or replaced by it M (t: x) to obtain

254

11.4.4 Cumulant Generation Function

11.4.5 Pearson Coefficent

Y, > 0 implies that exponential distribution is postively skewed and y2 > O means the
distribution is leptokurtic.

11.5 Summary
In this lesson, we have learnt the normal distribution and exponential distribution.
 255

11.6 Check Your Answers

I. a. bell b. coincide

11.7 Model Questions

1. X is normally distributed with mean and variance as 12 and 16 respectively. Find out the
probability of the following.

1. X > 20 2. X < 20 and 3. 0 < x < 12

Ans. i) 0.0228, ii) 0.9772, iii) 0.4987

2. The following tables gives frequencies of occurence of a variable X with in certain limits.

Variable X Frequency
Less than 40 20
40 or more but less than 50 32
50 or more 37

The distribution is exactly normal. Find the distribution and also obtain the frequency
between x = 50 and x = 60.

Ans. = 46.1  = 116.26 approx.

3. In a certain examination the percentage of passes and distinction were 46 and 9

respectively. Estimate the average marks obtained by the candidates, the minimum
pass and distinction marks being 40 and 75 respectively. Assume the distribution of
marks to be normal

Ans. = 36.4  = 28.2

4. 500 candidates in an examination were grouped in to 3 classes, I, II and III in descending

order of merit. The numbers in the first two classes were 46 and 132 respectively. The
lowest mark in class I is 60 and that in classes II is 50. Assuming the distribution to be
normal, find the average standard deviation of marks. Show that the 50% of the candidates
obtained marks between 39 and 53.

Ans. = 46.14  = 10.42

5. If X is Exp(r) with P(x < 1) = P(X > 1) Find var(X)

6. If X has exponential distribution with mean 2. Find P(X < 1/X < 2).
256

LESSON - 12
CORRELATION AND REGRESSION
Structure

12.1 Introduction

12.2 Learning Objectives

12.3 Scatter Diagram

12.4 Karl Pearson Correlation Coefficient

12.5 Some Basic Results on Correlation Coefficient

12.6 Correlation Coefficient for a Bivariate Frequency Distribution

12.7 Rank Correlation

12.8 Lines of Regression

12.9 Equation to the Regression Lines.

12.10 Angle Between two Lines of Regression

12.11 Summary

12.12 Check Your Answers

12.13 Model Questions

12.1 Introduction
In the previous lesson, we have mainly dealt with univariate universes only. In this lesson
we shall study about bivariate universes and in particular the study of simultaneous variation
of two variables X and Y (say). The variables may be amount of fertiliser applied and the yield
or the height and weight of the students in a class or the price and demand of a commodity. The
relationship between variables may be of any type but the linear type relationship is of much
importance.
 257

12.2 Learning Objectives

After studying this lesson, you should be able to

• understand the correlation coefficient and its limits

• calculate the correlation coefficient for bivariate data given in series

• calculate the correlation coefficient for data given in a two-way frequency table calculate
the rank correlation coefficient

12.3 Scatter Diagram

For a bivariate distribution (x, y) i = 1, 2..n, if the values of the variables X and Y are
plotted along the X-axis and Y-axis respectively in the xy plane, in Fig 12.1 the diagram of dots
obtained is called a scatter diagram. The points of the scatter diagram suggest whether the two
variables are correlated or not. If the points of the scatter diagram are very close to each other,
it suggest that there is a good amount of correlation between the variable. If they disbursed
widely, then it indicates lack of correlation.

Fig.12.1 Scatter diagram

258

12.4 Karl Pearson Correlation Coefficient

We know that E(X-X)2 measures the variance in X and E(Y-Y)2 variation in Y.

The simultaneous variation in X and Y is measured by E(X-X) (Y-Y). This measure in

terms of unit of measurement of X and Y. To make the independent of unit of measurement we
divide it by the quantity which has the same dimension as E(X-X) (Y-Y) Karl Pearson defined a
measure for correlation coefficient of X and Yitelem00 401

This is denoted by rxy and is called the product moment correlation coefficient or simply
the correlation coefficient. Since , and , are positive, the sign of the correlation coefficient is
the same as the sign of the product moment .

12.5 Some Basic Results on Correlation

12.5.1 Change of Origin and Scale

Let u and v be two variates defined by

 259

So, if h and K are both positive, then we have

Thus the correlation coefficient of two variables is independent of origin and the scale of
measurement. Because of this property, while calculating the value for r, we can choose any
convenient origin and scale.

12.5.2 Limits of Correlation Coefficient

We shall prove that the correlation coefficient varies between +1 and -1 i.e. LrJd” 1 or -
1d”rd”1

12.5.3 Schwarz’s Inequality

For any random variables X and Y [E(XY)]2 <E[X2]E[Y2]

Proof

For every real constant , the variate (X - Y)2 is always non-negative and therefore has
a non-negative expectation.

i.e., E(X - Y)2 > 0

or 2E(X2) - 2E(XY) + E(Y)2 > 0 for every real a.

In particular this holds for  = E(XY) / E(X2)

Substituting the value of a, we have

260

Proof of Theorem

If X’, Y’ denote deviation of the variates X and Y from their respective mean by the lemma,
we have,

Problem 1

Calculate the coefficient or correlation for the following distribution.

X 21 23 30 54 57 58 72 78 87 90

y 60 71 78 83 110 84 100 92 113 135

Solution
 261

Another Method

Since the correlation coefficient is unaltered by change of origin and scale, taking origin
A and B in X and Y, we have the formula for ras

Taking A as 57 and B as 84
262

Problem 2

With the usual notation derive the formula

Using the formula calculate the correlation coefficient between x and

y from the following table

X 11 13 20 44 47 48 62 68 78 80

Y 70 81 82 93 120 94 110 102 123 145

Solution
263
264

Problem 3

If X and Y are independent variables with mean 5 and 10 and standard deviations 2 and
3 respectively.

Obtain r(U, V), where U = 3X + 4Y and V = 3X-Y.

Solution

Since X and Y are independent Cov(X, Y) = 0

 265

12.6 Correlation Coefficient for a Bivariate Frequency Distribution

When the data is large, using a two way table, we can represent the data as a frequency
distribution. As in the case of univariate frequency distribution, choose a suitable class interval
for each of the variables.

If there are n classes for X and m classes for Y, then there will be m x n cells. From the
data for X and Y we can find the frequency for each cell. The column total and row total gives
the frequency for the two variables called the marginal distribution of the respective variable.

Problem 4

We shall illustrate the calculation of correlation coefficient for the bivariate table which
gives the percentage of marks in English and Mathematics for 300 students.

Marks in English

Marks in Maths 15-30 30-45 45-60 60-75

0-20 3 33

20-40 12 39 6

40-60 9 75 15 6

60-80 24 39 9

80-100 6 18 6
266

Solution

Let x, y represent the mark in England and Maths

Procedure

Let X denote the marks in English and Y denote marks in Mathematics. Write down the
middle values of x and y in the first row and first column. f, f, denote frequencies for x and y.
Take one of the mid-values of x and one of the mid-values of y as origins A and B. Here A = 37.5
and B = 50. Find the deviations of the other mid-values from these origins expressed in class
units.

First mid x being 22.5, the corresponding deviation X is

22.5-37.5 / 15 =-1
 267

Similarly the other values for X are 0, 1, 2. For the y series.

The entries under y are -2, -1, 0, 1, 2 respectively. Find f X, fX2, f,Y, f,Y2. Consider the
first cell frequency. Its X value from the first column is -1 and Y values from the first row is -2,
multiply these two and write in the right hand corner of the cell. Similarly do for all the other cells
having frequency. Multiply the cell frequencies with the corresponding XY values. Add these
products of the first column 3x2+12x1+9x0 = 18 and write it in fXY. Similarly do for other columns.
Repeat the same method adding row wise and put the same in the column with heading fYX.
Evidently the total of fXY and fYX will be the same. The formula for

Problem 5

From the following distribution of (x, y) find the correlation coefficient between x and y
268

Solution
 269

Check Your Progress - I

Fill in the blanks

a. Positive correlation implies that on an average as one variable is increasing the other is
________________and as one variable is decreasing the other is also_____________

b. If r = 0.03, r2 will

C. The coefficient of correlation is independent of change of ________________ and

_______________________

d. If r is negative both the variables are decreasing. T/F

12.7 Rank Correlation

Suppose we have a group of n individual ranked according to different characters, we
are interested to find the correlation between the rank which will reflect the correlation between
the two characters. Rank correlation is denoted by p.

The ranks of n individual for the two characters (x and y) will be the numbers 1, 2,...n in
some order.
270

Similarly,

If d, stands for the difference in ranks of the ith individuals, we have d1 = x,-y,,

The formula for rank correlation is due to Spearman.

Problem 6

Find the correlation between the rank in subject A and the ranks in another subject B
from the following data of marks in two subjects obtained by a class of 20 sheets.
 271

Solution

A 99 47 68 57 76 31 30

44 82 58 49 65 63 69

72 37 59 77 53 40

B 73 61 70 40 72 39 18

62 66 65 60 91 67 34

51 48 53 82 76 17
272

12.7.1 Repeated Ranks

In the case when two or more observations are repeated in any of the series, the common
ranks are given to repeated items. This common rank is the average of the ranks which these
items would have assumed if they were slightly different from each other and the next item will
get the rank next to the rank already assumed.

In the formula for Spearman’s rank correlation coefficient add the factory

m(m2-1)/12 to d2

where m is the number of times an item is repeated. The correction factor is to be

added to each repeated value.

Problem 7

A sample of 12 factors and their eldest sons gave the following data about their height in
inches. Calculate the rank correlation coefficient.

Father’s 65 63 67 64 68 62 70 66 68 67 69 71
Height

Son’s
Height 68 66 68 65 69 66 68 65 71 67 68 70
 273

Solution

X-Rank of Y-Rank of d=X-Y d2

father’s height son’s height

9 5.5 3.5 12.25

11 9.5 1.5 2.25

6.5 5.5 1.0 1.00

10 11.5 -1.5 2.25

4.5 3 1.5 2.25

12 9.5 2.5 6.25

2 5.5 -3.5 12.25

8 11.5 -3.5 12.25

4.5 1 3.5 12.25

6.5 8 -1.5 2.25

3 5.5 2.5 6.25

1 2 -1.0 1.00

72.50

In the first series there are two 68. Had they been slightly different, their ranks would
have been 6 and 7. So allocate the average 6.5 as rank. Similarly do for all repeated observation.
274

Check Your Progress - II

Say True or False

a. Rank correlation coefficient is denote by RTIF

b. The rank correlation coefficient was developed by spearman. T/F

C. If the values of x variables are 1, 2, 3, 4, 5 and those of y 4, 6, 8, 10, 12 the Karl Pearson
and the Rank method would give the same answer. T/F

12.8 Lines of Regression

In a bivariate frequency or probability distribution, if there exists association or relationship
between two variables X and Y, the points of the scatter diagram will be more or less concentrated
along a curve which we call as the curve of regression and the relationship is said to be expressed
by means of curvilinear regression. When the curve is straight line, it is called the line of
regression and the regression is said to be linear.

The line of regression is the line which gives the best estimate to the value of one
variable for a given value of the other variable in the least square sense. If the straight is so
chosen that the sum of squares of deviations parallel to y-axis is minimised, it is called regression
line of y on x and it gives the best estimate of y for any given value of x. If the sum of the squares
of deviations parallel to the x-axis is minimised the resulting straight line is known as the
regression of x on y. This gives the best estimate of x for any given value of y Thus for a given
distribution we have to regression lines.
 275

Fig. 12.2 Regression lines

12.9 Equation to the Regression Lines: Regression Coefficient

Let us suppose that in the bivariate distribution (x,, y,): i = 1, 2 .. n, Y is dependent
variable and X is the independent variable. The line of regression of Y on X is y = a + bx. The
expected value of Y for a given X = x, is y, = a + bx, and if S is the sum of the squares of
deviations of the observed value y,, for the expected value Y, then S is given by

According to the principle of least squares, the constant a and b are to be choosen so
that S is minimum. The normal equations are given by
276

multiplying (3) by and subtracting from (4), we get

So the regression line of y on x is

Similarly the regression line of x on y is

It is clear that both the regression lines passes through the mean (x, y). The coefficients.

are respectively called the regression coefficient of y on x and regression coefficient of

x aon y and they are usually denoted by by and by, respectively. Using the definition of

i.e the correlation coefficient is the geometric mean of the regression coefficients. The sign of

r is same as that of any regression coefficients. Since are always positive ,

and r have the same sign
 277

i.e. if one of the regression coefficient is less than 1, the other will be greater than 1 or both may
be equal to 1. The regression coefficient are independent of the change of origin but not of
scale.

r is a number free from unit of measurement, where as by, and by, are expressed in the
unit in which the variables are measured.

Problem 8

Calculate the coefficient of correlation from the following data.

X 1 2 3 4 5 6 7 8 9

Y 9 8 10 12 11 13 14 16 15

Also obtain the equations of the lines of regression and obtain an estimate of Y which
correspond to X = 6.2.
278

Solution

Equation of regression line y on x is

 279

Equation or regression line x on y is

Problem 9

The equations of two regression lines obtained in a correlation analysis are as follows

3x+12y=19, 3y+9x=46

Obtain (1) the value of correlation coefficient (2) the mean values of x and y (3) find
y when x = 1000.

Solution

Let 3x + 12y = 19 be the regression equation of y on x. then it can be written as

12y = 3x + 19 or

The coefficient of x is the regression coefficient of y on x. Hence byx = 1/4

Then 3y+9x=46 will be the regression line of x on y. It can be written as

280

Since the product of the regression coefficients = r2,

we get r2 = 1 /12

Since the sign of regression coefficient are negative, the sign of r is negative

ii) Since the regression lines passes through (x, y)

Solving the two equations

3x + 12y = 19 (1)

9x + 3y = 46 (2)

(2) 9x+3y=46

(1) x 3 9x+36y= 57

Subtracting -33y=-11

Substituting in (1) 3x + 4 = 1

3x = 15

i.e.x = 5

X = 5
 281

Problem 10

The following table gives the distribution of total cultivable area (x) and the area under
cultivation (y) in a district in 69 villages.

1. Find the regression line of y on x

2. The correlation coefficient rx, y.

3. The average area under wheat corresponding to total area of 1000 Bighas.

Area under Total area in Bhigas (x)

wheat (v) 0-500 500-1000 1000-1500 1500-2000 2000-2500

0-200 12 6 - - -

200-400 2 18 4 2 1

400-600 - 4 7 3 -

600-800 - 1 1 2 1

800-1000 - - 1 2 3
282
 283

Regression line of y on x is
284

i.e. y = 0.306x + 65.529

When x = 1000 y = 306 + 65.529

= 371.528

12.10 Angle Between Two Lines of Regression

Equation of regression lines of y on x and x on y are respectively

The slopes of the two lines are and respectively. If 0 is the acute angle

between the two regression lines, then,

 285

i.e. The lines of regression are perpendicular

i.e. the lines of regression are parallel or coincident.

But since the regression lines passes through (x, y) the two lines are coincident. Hence
in the case of perfect, positive or negative correlation, the two regression lines coincide.

Problem 11

Out of the two lines of regression given x + 2y - 5 = 0 and 2x + 3y-8=0, which one is the
regression line of x on y. Use the equation to find the mean of x and y. Calculate the variance of
y if the variance of x is 12.

Solution

Let us assume that x+2y-5=0 is the equation of regression line of y on x. Then the equation
can be written as

Hence the regression coefficient byx = - 1/2

Let the equation 2x + 3y-8=0 represent the equation of regression line of x on y.

Then the equation can be written as

Hence the regression coeffcient bxy = - 3/2

We know that the product of regression coefficient

286

Since byxbxy are of negative sign r=-

Since r<1 our assumption is correct. Had we assumed wrongly rX’ would be greate
than 1.

Solving

x + 2y = 5 ...(1)

2x + 3y = 8 ...(2)

(1) x 2 2x + 4y = 10

Subtracting -y=-2 i.e. y = 2

Substituting in (1), x+4= 5 x = 1 i.e. x=1

Ratio of regression coefficients = byx / bxy

Check Your Progress - III

Fill in the blanks

a. The regression coefficient of y on x is denoted by the symbol

b. If both the regression coefficient are negative the correlation coefficient would be

C. If two regression coefficients are 0.8 and 0.6 what would be the coefficient of correlation

12.11 Summary
In this lesson we have learnt the correlation coefficient, bivariate method, Rank correlation
procedure and also to find the equations of regression lines from a given data and to estimate
the values of one given by the other.
 287

12.12 Check Your Answers

I. a. Increasing, Decreasing

b. 0.09

c. Scale, Origin

II. a. F

b. T

c. T

III. a. byx

b. Negative

c. 0.693

12.13 Model Questions

1. Calculate the coefficient of correlation between birth rates and death rates from
the following table.

Year 1931 1932 1933 1934 1935 1936 1937

Birth rate 18.0 17.4 16.6 17.2 16.9 16.7 17.1

Death rate 11.1 10.9 10.7 11.1 10.9 11.6 11.3

Year 1938 1939 1940 1941 1942 1943 1944 1945

Birth rate 17.6 17.3 17.9 18.9 20.9 21.5 20.2 19.6

Death rate 10.6 10.6 10.7 10.5 10.4 10.9 10.6 10.6

Ans. r = -0.48

2. The following make have been obtained by a class of students in statistics.

Paper I 45 55 56 58 60 65 68 70 75 80 85

Paper II 56 50 48 60 62 64 65 70 74 82 90

Compute the coefficient of correlation for the above data.

288

3. Twelve students take tests in two subjects A and B with the following scores.

A 47 51 58 60 63 65 71 78 79 81 86 92

B 44 63 45 71 58 62 71 80 63 75 57 74

Determine the correlation coefficient.

4. Calculate the correlation coefficient between x and y where

X 9 10 11 12 13 14 15 16 17 18

Y 18 17 19 21 20 23 22 25 27 26

5. Find the correlation coefficient between age and salary of 50 workers in a factory from
the data given below.

Monthly pay in Rupees

Age 160-16 170-17 180-18 190-19 200-20 Total

9 9 9 9 9

20-30 5 3 1 9

30-40 2 6 2 1 11

40-50 1 2 4 2 2 11

50-60 1 3 6 2 12

60-70 1 1 5 7

60-70 1 1 5 7

Total 8 12 11 10 9 50
 289

6. Calculate the correlation of coefficient from the following table giving the ages of
100 husbands and their wives in years.

Age of Husbands

Age of 20-30 30-40 40-50 50-60 60-70 Total

wives

15-25 5 9 3 - - 17

25-35 - 10 25 2 - 37

35-45 - 1 12 2 1 15

45-55 - - 4 16 5 25

55-65 - - - 4 6 10

7. The ranks of the same 16 students in two subjects A and B were as follows.

Rank in A 1 2 3 4 5 6 7 8

Rank in B 1 10 3 4 5 7 2 6

Rank in A 9 10 11 12 13 14 15 16

Rank in B 8 11 15 9 14 12 16 13

Calculate the rank correlation coefficient.

8. Obtain the rank correlation coefficient for the following data

X 68 64 75 50 64 80 75 40 55 64

Y 62 58 68 45 81 60 68 48 50 70

Ans. 0.545
290

9. Compute the coefficient of correlation on the basis of ranking for the following data

X y

100 400

200 600

300 700

400 500

500 100

600 300

700 200

10. Why do we have in general two lines of regression? Obtain the line of regression of y on
x on y from the following take and estimate the blood pressure when the age is 50 years.

Age in years (x) 56 42 72 36 63 47 55 49

Blood pressure 147 125 160 118 149 128 150 145

Age in years 38 42 68 60

Blood pressure 115 140 152 155

11. The regression lines of x and y are given by 3x - 10y +4 = 0 and 10x - 27+82=0 and the
variance of x is 9. Obtain the means of x and y and the coefficient of correlation between
x and y.

12. Prove index numbers of wheat (X) and cereals (Y) at twelve successive seasons are
given below.

X 87 84 88 102 101 84 72 84 83 98 97 100

Y 88 79 83 97 96 90 82 84 88 100 80 102

a) Fit the line of regression of Y on X.

b) Suggest the value of Y when X is expected to be 110.

13. The following marks have been obtained by a class of students in Statistics

Paper I 45 55 56 58 60 65 68 70 75 80 85

Paper II 56 50 48 60 62 64 65 70 74 82 90

Compute the coefficient of correlation for the above data. Find also the equation
of regression lines.
 291

LESSON - 13
SAMPLING AND LARGE SAMPLE TESTS
Structure

13.1 Introduction

13.2 Learning Objectives

13.3 Parameter and Statistics

13.4 Sampling Distribution

13.5 Stand and Error

13.6 Test of Significance

13.7 Large Sample Test

13.8 Summary

13.9 Check Your Answers

13.10 Model Questions

13.1 Introduction
In statistics the word population or universe refers to aggregate of individual or their
attributes or of results of operations which can be expressed numerically. For example, we
may say population of weight, population of height etc. The population may be finite or infinite. A
population containing finite number of members is called finite population. The population with
infinite number of members is called an infinite population.

In many statistical investigations complete enumeration of population is impractical.

For example, if we are interested to find the average per capita income of the people in India, we
have to enumerate all the earning individual in the country which is rather difficult task.

13.2 Learning Objectives

After studying this lesson you should be able to understand
292

• the terms sampling distribution, standard error.

• test of signficance large sample test.

13.3 Parameter and Statistics

If the population is infinite, complete enumeration is not possible. Even if the population
is finite, enumeration is possible only in certain cases. For example, if we want to find the
average burning hours of a bulb produced by a factory, we cannot burn each and every buld till
it fuses, to get the average lifetime of the bulb. In that case on buld will be available for consumption
by public. Even if complete investigation is possible we resort to sampling method due to
administrative, financial implications and tme factors. By sample, we mean a part or small or
small portion of the population and the process of such selection is known as sampling. To
serve useful purpose the sample should be true representative of the population.

In sampling theory our aim is to get as much information as possible about a population
from the samples drawn. By knowning the form of population, we an estimate the parameters
(i.e. statistical constants like mean, standard deviation, etc.) of the population or we may find
the limit within which the population parameters lie under certain level of probability. The statistical
constants of the samples are referred as statistics. Since the size (number of observations) of
the sample is small there is bound to be an error and this error is known as sampling error.

In practical life we resort to sampling quite often. For example if we want to purchase a
bag of rice, we take a handful of rice at random from the bag and examine about the quality and
then decide to purchase. A sample is called a random sample if each individual of the population
have the same chance of being selected.

13.4 Sampling Distribution

Consider a finite population of size N, we can draw NCn samples from the population of
size n. For example if the size of the population is 100 and if we require samples of size 4, we
can draw
 293

For each of the samples we can compute a statistic, say mean (x) or standard variance
(S2). The value will vary from sample to sample. The collection of all such values of the statistics
is known as the sampling distribution of the statistic.

13.5 Standard Error

The standard deviation of the sampling distribution of a statistic is known as standard
error. If we consider the sampling distribution of the statistic mean than its standard deviationis
known as standard error of mean.

13.5.1 Standard Error of a Sample Mean

The standard error of the mean of random sample of size n from a population

with variance

Let x1, x2 ............xn be a random sample of size n from a population with variance 2,
then the sample mean x is given by

But the covariance term vanishes since the sample observations are independent). We
have

Var(x) = E[x, - E(x)]2 = E(x, - u)2

=1/n [(x,- u)2 + (x2 - u)2 + .. +(xn - u)2] = 2 i = 1, 2..n

294

i.e., Standard error of

Similarly SE of observed sample proportion p is

S.E. of the standard deviation s is

Standard errors play an important role in the large sample theory and forms the basis of
testing of hypothesis. If t is any statistic then for large samples

follows the normal distribution with mean zero and unit variance i.e.,

Check Your Progress - I

Fill in the blanks

a. Sample mean is denoted as ____________________

b. _____________________

13.6 Test of Significance

An important aspects of the sampling theory is the study of the test of significance. This
enables us to decide on the basis of the sample results whether

i) The derivation between the observed sample statistic and the hypothetical population
parameter and
 295

ii) The derivation between two sample statistics is significant or may be attributed due to
chance or fluctuations in sampling.

For large n, almost all distribution like bionmial, Poisson, negative binomial,
hypergeometric, etc. can be approximated very closely to a normal probability curve. Hence we
use the normal test of significance for large sample. Some well-known test of significance for
small sample are t, F, x2 test and Fisher’s transformation.

13.6.1 Null Hypothesis

A hypothesis is a definite statement about the population parameter. Such hypothesis

which is usually a hypothesis of no difference is called Null hypothesis and is usually denoted
by Ho. According to Prof. [Link], null hypothesis is the hypothesis which is tested for
possible rejection under the assumption that it is true.

Complementary hypothesis to the null hypothesis H, is called the alternative hypothesis

usually denoted by H1.

For example if

13.6.2 Errors in sampling

The main objective of the sampling theory is to draw valid inference about the population
parameters on the sample results. We accept or reject a hypothesis based on the values
obtain from a sample. As such, we are likely to commit two types of erros.

Type I Error

To reject the hypothesis H. when it is true.

Type II Error

To accept Ho. when it is wrong.

i.e. to accept Ho. when H, is true.

If we write, P(reject Ho, when it is true) = P(reject Ho/Ho) = 

P(accept Ho when it is wrong) = P(accept Ho/H1) = .

Then a and  are called the sizes of Type I Error and type II Error respectively.
296

13.6.3 Critical Region and Level of Signficance

A region (corresponding to a statistic) in the sample space which amounts to the rejection
of Ho is called critical region or region of rejection. If  is the critical region and if (X,, X,.....X) is
the random vector observed, then P(X € /Ao) = , P(X €  /H1) - , where is the complementary
of  called region of acceptance. The maximum probability that a random value of the vector X
= (X1, X2...Xn) belong to the critical region is called the level of significance. In testing hypothesis,
we usually adopt 5% and 1% level of significance.

13.7 Large Sample Test

Let Z be a standardised normal variate and

P[IZ 1.96] = 0.05, then

We say that Ho is rejected under 5% level. If |Z| 2.98, Ho is rejected under 1% level of
signficance.

13.7.1 Test Based on Sample Mean poed

We are interested to find whether a random sample with mean is from a population with
mean X. The null hypothesis to be tested is Ho: = against H,  -

Let and X be the mean and standard deviation of a large sample (size n > 30). We know
that the standard error of

The test statistic

We reject H, if |Z| > 1.96 under 5% level of signficiance.

 297

Problem

A sample of 1600 members is found to have a mean of 3.2cms. Could it be resonable

regarded as a sample from a large population whose mean is 3.00cms and standard deviation
2.61cms.?

Solution

Given the sample size n = 1600

Sample mean X = 3.2 cm

Population mean  = 3.0 cm

Population S.D.  = 2.61 cm 30%

We shall have the null hypothesis that the given sample with mean = 3.2 cm is a sample
from population with mean ì = 3.0 cm and S.D. = 2.61

We shall take 5% level of significance, then

Since Z 1.96. We reject the null hypothesis under 5% level of significance. The sample
may not be reasonably considered to be from the population.

Problem 2

A sample of 144 men gave the mean height of 5’5" with a standard deviation of 2.5".
Discuss the suggestion that the sample does not form a part of the population whose mean
height is 5'. 6".

Solution

The size of the sample n = 144.

Sample mean x = 5’5"

Sample standard deviation S.D. = 2.5"

298

We shall have the null hypothesis that the simple with mean 5’5" and standard
deviation 2.5" is from the population with mean 5’6". The statistic

since n is a large

Since |Z| > 0.96, the difference between the two means is significant under 5% level of
significance.

There is every chance for the sample not to be from the population.

13.7.2 To test the Significance of the Difference Between the Means of two
Samples

Let x1 and x2 be the means of two samples of size n1 and n2 from a population with a
standard deviation ó. We shall have the null hypothesis that the samples are from the same
population. The standard [Link] of the difference between the mean is

So the statistic

If |Z| > 1.96, then we reject the null hypothesis that the samples are from the same
population under 5% level of signficance. If |Z| < 1.96 we accept the null hypothesis.

Note: If instead of the standard deviation of the population, the standard deviation of the
samples s, and s, are known, an estimate of o based on s1 and s2 is

Since the sample is large n1 + n2- 2~n1 + n2

 299

Problem 3

A certain intelligence test administered to a large group of pupils gave the standard
deviation of scores as 30. The test is given to group of 200 boys who got an average score of
128. Another group of 150 girls who took the same test scored an average of 135. Does this
show any significant difference between the two group?

Solution

With the usual notation, we are given that

Under the null hypothesis Ho: 1 2

i.e. the average scores of the two populations are equal

The statistics Z =

Here Z =

Since z> 1.96 the error between the two sample means is significant under 5% level.

Problem 4

Two groups of 500 men and 1500 men are found to have mean heights 68.6" and 66.7"
with the corresponding variances 6.2 and 5.9 square inches. Is the difference of means
significant?
300

Solution

under the null hypothesis , the statistic

Since |z| > 1.96 we reject the null hypothesis under 5% level of significance. The difference
between the means of the two groups is highly significant.

13.7.3 Test for Single Proportion

If X is the number of success in n independent trails with constant probability p of success

for each trial then E (X) = np, Var(X) = npq when q = 1-p. For large n, the Binomial distribution
tends to Normal distribution. Hence for large n,

X - N(np, npq)

Hence Z

and we can apply the normal test

Problem 5

A die is thrown 9000 times and a throw of 3 or 4 is observed for 3240 times. Show that
the die cannot be regarded as an unbiased one.

Solution

If getting 3 or 4 is considered as sucess, n = 9000 X = the number of success 3240

Under the null hypothesis Ho: that the die is an unbiased one we get probability of success
 301

Hence Z

Since z> 3, H is rejected under 1% level of significance. We conclude that the die is
almost biased.

13.7.4 Test of Significance of the Difference of Proportions

Let x1, x2, be the number of persons possessing a certain attribute A in random samples
of sizes n, and n, from two populations respectively. Then the sample proportions are

Under the Null Hypothesis Ho: P1 = P2 i.e., there is no significant difference in the
population proportions, we have,
302

since under Ho: p1 = P2 =P (say); Q1 = Q2 = Q.

Hence under Ho the statistic

P is estimated as

Note: If the sample proportion are not given, we setup the null hypothesis p, P, Hence
the test statistic in this case becomes

Problem 6

In a sample of 400 men in a certain city 250 are found to be smokers. In another city out
of 800 men 350 are found to be smokers. Do the data indicate that the cities are significantly
different with respect to the prevalence of smoking among men?

Solution

The null hypothesis H: P, P, i.e. the cities are not significantly different with respect to the
prevalence of smoking among men.

With usual notation

 303

Since Z> 1.96, HO is rejected under 5% level. There is significant difference between the
two cities regarding the prevalence of smoking among men.

Check Your Progress - II

Say True or False

a. Sample size > 30. We use the large sample T/F.

b. If |z| > 2.98, Ho is rejected under 5% level of significant T/F.

304

13.8 Summary
In this lesson we have learnt the sampling distribution, standard error, and large sample
tests.

13.9 Check Your Answers

I. a. x II. a. T

b. 0,1 b. F

13.10 Model Questions

1. A sample of 900 members has a mean 3.4cm and standard deviation 2.61cms. Is the
sample from a large population of mean 3.25cm and standard deviation 2.61cm.?

2. The means of two large sample of 1000 and 2000 members are 6.75 and 68.0 inches
respectively. Can the samples be regarded as the sample drawn from the sample
population of standard deviation 2.5 inches.

3. In a survey of buying habits of 400 women shoppers are chosen at randomin a

super market A located in a certain section of the city. Their average weekly food
expenditure is Rs 250 with a standard deviation of Rs 40. For 400 women shoppers
chosen at random in super market B, in another section of the city, the average weekly
food expenditure is Rs 220 with a standard deviation of Rs 55. Test at 1% level of
significance whether the average weekly expenditure of the two populations are equal.

4. A coin is tossed 10000 times and it turns up head 5195 times. Discuss whether coin may
be considered as an unbiased one.

5. In a large consignment of oranges a random sample of [Link] revealed that

14 oranges were bad. List reasonable to assume that 20% oranges were bad?

6. From each of two consignment of apples, a sample of size 200 is drawn, and the number
of rotten apples counted. Test whether the proportion of rotten apples in the two
consignments are significantly different or not, given that

Size of Sample No. of Rotten Apples

Sample from consignment A 20 44

Sample from consignment B 20 30

7. A machine puts out 16 imperfect articles in a sample of 500. After the machine is over
handed, it puts out 3 imperfect articles in a batch of 100. Has the machine improved?
 305

LESSON - 14
SMALL SAMPLE TESTS
Structure

14.1 Introduction

14.2 Learning Objectives

14.3 Student t Definition

14.4 Chi-Square Variate Definition

14.5 Chi-Square Test of Independence of Attributes

14.6 Summary

14.7 Check Your Answers

14.8 Model Questions

14.1 Introduction
If the sample size n is small, we can not apply test based on normal distribution. For
such cases, the t test, F test and Fisher’s Z transformation are applicable. In all these test the
basic assumption is that the population from which the samples are drawn is normal. Here we
shall discuss certain applications of t distribution.

14.2 Learning Objectives

After reading this lesson you should be able to understand

• tests based on t and 2 distributions

• to do problems of test of significance (small sample)

• test goodness of fit and independence of attributes

14.3 Student t Definition

Let x, (i = 1, 2...n) be a random sample of size n from a normal population with mean m
and variance o2. Then the Student’s is defined by the statistic
306

Ex - the sample mean

and S2 =

is an unbiased estimate of the population variance o2 and it follows the student’s t

distribution with degrees of freedom u = (n-1) with probability density functiona

The t distribution has a wide number of applications. Here we consider two tests.

1. To test the significance of the difference between sample mean and the population mean

2. To test the significance of difference between two sample means.

14.3.1 Test for Single Mean

Let X be the mean of a sample of size n from a normal population whose mean is .
Under null hypothesis HO. There is no significant difference between the sample mean and the
population mean , are statistic

follows the t distribution

with degrees of freedom v = (n - 1).

Here

After calculating the value of t from the given data, we compare this value of t with the
tabulated value of t at certain level of significance and if the calculated value of t< tabulated
value of t. His acceptance at the level of significance adopted.
 307

Problem 1

An independent sample of 8 items had the following values of the variables (weight in
Kgs.) 8, 12, 11, 14, 13, 15, 13. Does the mean wight of the sample differe significantly from the
population mean of 12 Kgs.

Solution

The sample mean

Mean square deviation about Zero origin is

Population mean ì = 12

Variance of the sample is S2 = 154-(12.25)2

= 3.9375

We know that ns2 = (n-1)S2

The test statistic become

Referring to the t table for degree of freedom = 8-1=7 and the probability P=0.05 the
value of t is 2.37.

Since t< 1.05 the difference between the sample mean and populatin mean is
insignificant.
308

14.3.2 t Test for difference of Means

Let x, (i = 1, 2...n,) and yj (j = 1, 2...n2) be two independent samples from two normal
population with means ì and ì, respectively. Under the null hypothesis that the samples have
been drawn from the same population and under the assumption that the population variances
are equal.

is known as unbaised estimate of the population variances o2 and follows the student’s
t distribution with degress of freedom v = (n, + n2-2).

An Important deduction

Under the null hypothesis that the samples have been drawn form population with same
mean ìy, then the statistic,

follows the student’s t distribution with degrees of freedom from (n, + n2 - 2).

Problem 2

Two independent samples of 10 and 8 items respectivley from the same population had
the following values of the variable (weight in Kgs.)

Sample I 4, 6, 9, 8, 10, 9, 8, 12, 14, 10

Sample II 6, 8, 9, 7, 8, 6, 8, 9

For I sample the mean square deviation about zero origin is 882/10 = 88.2

Variance s21 = 88.2-(81) = 1.200

 309

For I sample the mean square deviation about zero origin is 475/8 = 59.375

Variance S22 = 59.375-(7.625)2

= 59.375-58.141

= 1.234

From t table for v = 10+8-2= 16 and for probability 0.05 t 0.5. = 2.12. Since the
calculated value t>t0.5, the difference between the two sample mean is significant.

14.3.3 Paried t Test for Difference of Means

Let us now consider the case when the sample sizes are equal. i.e. n, = n, = n and the
samples are not independent but the samples observations are paired together. The problem
is to test if the sample means signficiantly differ or not.

Let the pairs be (x1, y1) (x2 y2) ........(xn yn).

Define U1 = (xi, yi) i = 1, 2..n and Su= standard deviation of u1, u2...un then the static
310

follows the t distribution with degrees of freedom v = n - 1..

Problem 3

Eleven school boys were given a test in Mathematics. They were given a months tuition
and a second test was held at the end of it. Do the marks give evidence that the student have
benefitted by the extra coaching.

Boys Marks in the 1st test Marks in the 2nd test

1 23 24

2 20 19

3 19 22

4 21 18

5 18 20

6 20 22

7 18 20

8 17 20

9 23 23

10 16 20

11 19 17
 311

Solution

x y u = (y-x)

23 24 1

20 19 -1

19 22 3

21 18 -3

18 20 2

20 22 2

18 20 2

17 20 3

23 23 0

16 20 4

19 17 -2

For d.f. u = 10 and P = 0.05, to= 2.23

The difference between the performance of the student’s before and after coaching is
not significant. Data shows that extra coaching has not benefitted the students.,

14.4 Definition Chi-square variate

The square of a standard normal variate is known as chi-square variate with degrees of
freedom 1.
312

In general if Xi (i = 1, 2..n) are n independent normal variates with mean and variance’
o2(i) = 1, 2...n, then

14.4.1 Chi-square test of goodness of fit

If 0,= (i = 1, 2...n) is a set of observed frequencies and e, = (i = 1, 2,...n) is

the corresponding set of expected frequencies, then Karl Pearson’s Chi-square is given by

follows the chi-square distribution with (n-1) d.f

Problem 4

A survey of 320 families with 5 children each revealed the following distribution.

No. of Boys 5 4 3 2 1 0

[Link] Girls 0 1 2 3 4 5

[Link] Families 14 56 110 88 40 12

Is the result consistent with the hypothesis that male and female births are
equally probable.

Solution

The null hypothesis be that there is equal probability for male and female birth. Under
the null hypothesis
313
314

The expected frequencies e1 and the observed frequencies O1 are given below:

O1 e1 (0,- e)2 (0 - e)2

14 10 16 1.6000

58 50 36 0.7200

110 100 100 1.0000

88 100 144 1.4400

40 50 100 2.0000

12 10 4 0.40000

320 320 7.1600

From table for d.f 6-1 = 5 is 11.07

since calculated value of the null hypothesis may be accepted.

14.5 - Test of Independence of Attributes

Let A and B be two attributes. A is divided into r classes A,, A,....A, and B is dividend into
s classes B, BB. The various cell frequencies are given in the following table known as rxs
contingency table where (A,) is the number of persons possessing attribute A (i = 1,...r) and (B)
is the number of persons possessing attribute B, (j = 1...) and (A, B) is the number of persons
possessing both attributes A, and B, (i = 1......r; (j = 1, 2,...s)

= the total frequency.

 315

The problem is to test whether the two attributes are independent or not.

Under the null hypothesis that the attributes are independent, the cell frequencies
are calculated as following.

P[A] - probability that a person possess the attribute A1

= (A1)/N

P[B1]= (B1) / N

P[A1 B2] = P[A 1] P[B 2] by compound probability theorem since the attributes are
independent under null hypothesis

The number of degrees of freedom for 2 is (r-1) (s-1).

(Note: If an expected frequency is <5, the it can be added to the next class and
correpondingly the difference is adjusted.)

Problem 5

From the following table regarding the eye colour of fathers and sons. Test if the eye
colour of sons is associated with that of father.
316

Eye colour of son

Not light Light Total

Not Light 230 148 378

Light 51 471 522

Total 281 619 900

Under the null hypothesis the eye colour of sons and eye colour of father are independent.
The expected frequencies are

Since calculated x2 is much greater than the tabulated value, the null hypothesis that
the eye colour of father and son are independent is rejected under 5% level of significance.

Check Your Progress - I

Say True or False

a. Large sample theory can be used for small sample theory T/F

b. Let x be the sample mean of size 8 and population mean m. the statistic t =

~ the student t distribution, with degress of freedom T\F

C. The conditions for the validity of x2 test is the sample observations should be independent
and n is large T/F
 317

14.6 Summary
In this lesson we have learnt small sample Techniques.

14.7 Check Your Answers

I. a. F

b. T

C. T

14.8 Model Questions

1. The following are the length in inches of 12 random samples of Egyptain cotton taken
froma large consignment.

48, 46, 49, 46, 52, 45, 43, 47, 47, 46, 47, 50. Test the hypothesis that the mean length of
the consignment is 46".

2. Below are given the gain in weight (in lbs) of pigs fed on two diets A and B.

Gain in weight

Diet A 25, 32, 30, 34, 24, 14, 32, 24, 30, 31, 35, 25

Diet B 44, 34, 22, 10, 47, 31, 40, 30, 32, 35, 18, 21, 35, 29, 22

Test if the two diets differ significantly as regards their effects on increase in weight.

3. The length of 5 piece of cloth before and after washing are given below. Examine whether
washing affects length.

Before washing 10.2 6.4 7.3 20.3 30

After washing 10.4 6.5 7.5 20.8 30.9

4. The frequencey distribution of the digits in a set of random members was observed to
be

Digit 0 1 2 3 4 5 6 7 8 9

Frequency 24 18 16 14 20 25 17 18 23 25
318

Test the hypothesis that the digits are uniformly distributed i.e., have the same theoretical
frequencies.

5. The following table gives the number of aircraft accident that occured during the various
days of the week. Test whether the accidents are uniformly distributed over the week.

Days Mon Tue Wed Thu Fri Sat

[Link] accidents 14 18 12 11 15 14

6. The table below gives the date obtained during an epidemic of cholera Inoculated

Attacked Not attacked Total

Inouculated 31 469 500

Not Inouculated 185 1,315 1,500

Test the effectiveness of inoculation in preventing of cholera.

7. In a locality 100 persons were randomly selected and asked about their
educational achievements. The results are as follows:

Edication

Sex Middle Matric B.A. Total

Male 10 15 25 50

Female 25 10 15 50

Total 35 15 40 100

Can you say that education depends on sex.

319
320
321
322

MODEL QUESTION PAPER

BACHELOR OF COMPUTER APPLICATIONS
Second Year - Third Semester
Core Paper - VII
NUMERICAL AND STATISTICAL METHODS
Time : 3 Hrs. Max marks 75

PART A ( 10  2  20 marks)

Answer any TEN questions

Each Question carries 2 Marks

1. Write the formula for False position method.

2. Explain Gauss elimination method.

3. What are the demerits of Mode.

3
4. State Simpson’s - rule.
8

5. Write second-order Runge-Kutta formula.

6. Write the uses of Ogive.

7. What is the coefficient of Range and coefficient of variation ?

8. State Baye’s Theorem.

9. Explain Mutually exclusive events with example.

10. Find the moment generating function of the Poisson Distribution.

11. Write the formula for Karl Pearson Correlation coefficient.

12. Write the formula for t-test for difference of means.

 323

PART B ( 5  5  25 marks)
Answer any FIVE questions
Each Question carries 5 Marks

13. Find the real root of the equation f  x   x 3  2 x  5  0 .

1.5
x2
14. Evaluate e
0.2
dx using Romberg integration correct to five decimal places.

dy 2x
15. Using Improved Euler’s Method, Find y at x  0.1 and x  0.2 given  y  , y 0  1
dx y

16. Calculate the mode for the following distribution

Class Limit Frequency

45-49 2

50-54 3

55-59 5

60-64 7

65-69 9

70-74 11

75-79 7

80-84 2

85-89 3

90-94 1

17. Derive Moment generating function of Binomial Distribution.

18. Assume that mean height of soldiers to be 172 cms and the variance 27 cms. How many soldiers
in a regiment of 1000 can be expected to be over 182 cms.

19. A certain intelligence test administered to a large group of pupils gave the standard deviation of
scores as 30. The test is given to a group of 200 boys who got an average score of 128. Another
group of 150m girls who took the same test scored an average of 135. Does this show any significant
difference between the two groups ?
324

PART C ( 3  10  30 marks)
Answer any THREE questions
Each Question carries 5 Marks

20. Solve the following system of equations by Gauss-Seidel method:

28x  4 y  z  32
x  3 y  10 z  24
2 x  17 y  4 z  3

21. By applying the fourth-order Runge-Kutta method find y 0.2  from y   y  x, y 0   2 taking

h  0. 1

22. Calculate the mean deviation about the mean for the following data

Class 0-10 10-20 20-30 30-40 40-50 50-60 60-70

Frequency 4 10 15 35 52 40 28

23. The standard deviation of a certain group of 100 high school grades was 11% and the mean grade

78%. Assuming the distribution to be normal, find

i. How many grades were above 90%

[Link] was the highest grade of the lowest 10 ?

iii. What was the inter-quartile range?

iv. Within what limits did the middle 90% lie?

24. A survey or 320 families with 5 children each revealed the following distribution.

No. of Boys 5 4 3 2 1 0

No. of Girls 0 1 2 3 4 5

No. of Families 14 56 110 88 40 12

Is the result consistent with the hypothesis that male and female births are equally probable.