Unit - IV

Sampling Distribution and Queuing Theory

Sampling theory:-Sampling, Sampling distributions, standard error, test of hypothesis for

means and proportions, confidence limits for means, student’s t-distribution, Chi-square
distribution as a test of goodness of fit.
Introduction to queuing systems: Single queuing nodes, service disciplines, M/M/1 Queuing
system.

Random sampling:

Population:

A large collection of individuals or attributes or numerical data can be understood as a

population or universe.

Sample:

A finite subset of the population is called a sample.

The number of individuals in a sample is called a sample size.

Note: If the sample size (n) is less than or equal to 30 the sample is said to be small,
otherwise it is a large sample.

The process of selecting a sample from the population is called as sampling.

The selection of an individual or item from the population in such a way that each has the
same chance of being selected is called as random sampling.

(Or) Random sampling is a technique in which each of the sample has an equal chance of
being selected.

Suppose we take a sample of size n from a finite population of size N, then we will
N
have Cn possible samples.

Sampling with replacement and without replacement:

Sampling where a member of the population may be selected more than once is called as
sampling with replacement.

If a member cannot be chosen more than once is called as sampling without replacement.

Simple sampling is a special case of random sampling in which trials are independent and the
probability of success is a constant.

The word statistic is often used for the random variable or for its values.

1
Sampling Distributions:

Let us suppose that we have different samples of size n drawn from a population. For each
and every sample of size n we can compute quantities like mean, standard deviation etc.
Obviously these will not be the same. The statistics are grouped according to their
frequencies and from a frequency distribution. The frequency distributions so generated are
called sampling distributions. These can be distinguished as sampling distributions of mean,
standard deviation etc.

The sampling distribution of large samples is assumed to be a normal distribution.

The standard deviation of a sampling distribution is also called the standard error (S.E.). The
reciprocal of the S.E. is called precision.

Sampling distribution of the means:

Consider a population of size N for which mean is and standard deviation is σ.

Let  x and  x be mean and standard deviation of sampling distribution of means of a set of
samples of size n.

Sample mean is a statistic and we discuss the sampling distribution of this static.
We consider all possible random samples of size n and determine the mean of each one of
these samples. We discuss the sampling distribution of the sample means for the two possible
types of random sampling (with/without replacement) associated with a finite population.

Case (i): Random sampling with replacement

Thus we have  x  

and  x   / n .

Case (ii): Random sampling without replacement

x   ,

 N n
x 
n N 1

2
Example: Let {1, 2, 3} constitute a population we form the sampling distribution of the
sample means in the case of (i) random sample of size 2 with replacement (ii) random sample
of size 2 without replacement.

Solution: Here N (size of the finite population) = 3

1 2  3
Population mean    2
3

1 2
Population variance   2 
3
(1  2)2  (2  2) 2  (3  2) 2  
3

Case (i) Random sample of size 2 (n = 2) with replacement

The various possible samples are

(1, 1), (1, 2), (1, 3), (2, 1), (2, 2), (2, 3), (3, 1), (3, 2), (3, 3)

These are N n  32  9 in number.

The mean of these are respectively 1, 1.5, 2, 1.5, 2, 2.5, 2, 2.5, 3.

Frequency distribution of means

x 1 1.5 2 2.5 3
f 1 2 3 2 1
We shall compute the mean and variance of this frequency distribution

x 
 fx  1  3  6  5  3  2
f 9

x 2

 f (x   x )2

1
1(1  2)2  2(1.5  2)2  3(2  2)2  2(2.5  2)2  1(3  2)2 
f 9

1 1
 1  0.5  0  0.5  1 
9 3

1 2 2 3 1
Thus we have  x  2   ,  x2  and   .
3 n 2 3

Case (ii) Random sample of size 2 without replacement:

We have 3C2  3 samples these samples are (1, 2), (2, 3), (3, 1) and the associated means are
1.5, 2.5, 2.

1.5  2.5  2
The mean of these is given by  x  2
3

3
 1 0.5 1
The variance of these is given by  x2 
3
(1.5  2) 2  (2.5  2) 2  (2  2) 2  
3

6

Here we note that  x  2  

 N  n 
2
3 2 2 / 3 1
Also        x2
 N 1  n  3 1  2 6

 N  n 
2
1
Thus  x2     .
 N 1  n 6

Example: A population consists of 4 numbers 3, 7, 11, 15

(a) Find the mean and standard deviation of the sampling distribution of means by
considering samplings of size 2 with replacement.

(b) If N, n denotes respectively the population size and sample size σ and  x respectively
denotes population standard deviation and standard deviation of sampling distribution of
means without replacement verify that

2  N n
(i)  x2    (ii)  x   where  x is mean of the sampling distribution and  is the
n  N 1 
population mean.

Solution: Here N (size of the finite population) = 4

3  7  11  15
Population mean    9
4

1
Population variance   2 
4
(3  9)2  (7  9)2  (11  9)2  (15  9) 2   20

(a) Random sample of size 2 (n = 2) with replacement

The various possible samples are

(3, 3), (3, 7), (3, 11), (3, 15),

(7, 3), (7, 7), (7, 11), (7, 15),

(11, 3), (11, 7), (11, 11), (11, 15),

(15, 3), (15, 7), (15, 11), (15, 15),

These are N n  42  16 in number.

The mean of these are respectively

3, 5, 7, 9,

4
5, 7, 9, 11,

3, 7, 9, 11,

13, 9, 11, 13, 15

Frequency distribution of means

x 3 5 7 9 11 13 15
f 1 2 3 4 3 2 1
We shall compute the mean and variance of this frequency distribution

x 
 fx  3  10  21  36  33  26  15  144  9
f 16 16

x 2

 f (x   1 x )2
1(3  9)2  2(5  9)2  3(7  9)2  4(9  9)2  3(11  9)2  2(13  9) 2  1(15  9) 2 

f 16
1 160
 36  32  12  0  12  32  36   10
16 16

Thus we have  x  9   ,  x  10 .

(b) Random sample of size 2 without replacement:

We have 4C2  6 samples these samples are (3, 7), (3, 11), (3, 15), (7, 11), (11, 15) and the
associated means are 5, 7, 9, 9, 11, 13.

Frequency distribution of sampling means is as follows

x 5 7 9 11 13
f 1 1 2 1 1

5  7  18  11  13 54
The mean of these is given by  x   9
6 6

Here we note that  x  9  

The variance of these is given by

1 40 20
 x2  1(5  9) 2  1(7  9) 2  2(9  9) 2  1(11  9) 2  1(13  9) 2   
6 6 3

 N  n 
2
4  2  20 20
Also         x2
 N 1  n  4 1  2 3

5
Example: The weights of 1500 ball bearings are normally distributed with a mean of 635
grams and standard deviation of 1.36 grams. If 300 random samples of size 36 are drawn
from this population determine the expected mean and standard deviation of sampling
distribution of means if sampling is done (i) with replacement (ii) without replacement.

Solution: Given N = 1500, n = 36, μ = 635, σ = 1.36

(i) with replacement

Expected mean  x    635

 1.36
Expected standard deviation =  x    0.227
n 36

(ii) without replacement

Expected mean  x    635

 N n
Expected standard deviation =  x   0.224
n N 1

Example: Certain tubes manufactured by a company have mean life time of 800 hrs and
standard deviation of 60 hrs. Find the probability that a random sample of 16 tubes taken
from the group will have a mean life time

(i) between 790hrs and 810 hrs (ii) less than 785 hrs (iii) more than 820 hrs (iv) between 770
and 830 hrs.

Solution: Given μ = 800, σ = 60, n = 16

 60
x    15
n 16

x
z
x

(i) P(790  x  810)  P(0.67  z  1.67)  2 P(0  z  0.67)  2 A(0.67)  0.49714

(ii) P ( x  785)  P( z  1)  P( z  1)  P( z  0)  P(0  z  1)  0.5  A(1)

= 0.5 - 0.34134 = 0.15866

(iii) P( x  820)  P ( z  1.33)  P( z  0)  P(0  z  1.33)  0.5  A(1.33)  0.5  0.4082

= 0.0918

(iv) P(770  x  830)  P(2  z  2)  2 P (0  z  2)  2 A(2)  0.9545

6
Testing of Hypothesis

In order to arrive at a decision regarding the population through a sample of the population
we have to make certain assumption referred to as hypothesis which may or may not be true.

Null Hypothesis:The hypothesis formulated for the purpose of its rejection under the
assumption that it is true is called the Null Hypothesis denoted by H0.

Alternate Hypothesis:Any hypothesis which is complimentary to the null hypothesis is

called Alternate Hypothesis denoted by H1.

Example:

(1) To test where a process B is better than a process A we can formulate the hypothesis as
there is no difference between the process A and B.

(2) To test whether there is a relationship between two variables we can formulate the
hypothesis as there is no relationship between them.

Errors:

Type I error: Instead of accepting the null hypothesis if we reject H0 then we say type I error
has occurred.

Type II error: Instead of rejecting the null hypothesis if we accept H0 then we say type II
error has occurred.

Significance level:

The probability level, below which leads to the rejection of the hypothesis is known as the
significance level. This probability is conventionally fixed at 0.05 or 0.01 i.e. 5% or 1%.
These are called significance levels. We feel confident in rejecting a Hypothesis at 1% level
of significance than 5% level of significance. 5% level of significance can also be understood
as, the probability of committing errors of either types (Type I or Type II) is 0.05.

Tests of Significance and Confidence Intervals:

The process which helps us to decide the acceptance or rejection of the hypothesis is called
the test of significance.

Let us suppose that we have a normal population with mean  and standard deviation. If x
is the sample mean of a random sample of size n the quantity z defined by

x 
z (1)
( / n )

is called the standard normal variate (S.N.V.)

7
From the table of normal areas we find that 95% of the area leis between z = - 1.96 and z = +
1.96. In other words we can say that 95% confidence that z lies between – 1.96 and + 1.96.
Further 5% level of significance is denoted by z0.05 .

Thus we can write the verbal statement in the mathematical form

x 
1.96   1.96
( / n )

1.96( / n )  x    1.96( / n )

   x  1.96( / n ) and x  1.96( / n )  

Thus we can write by combining the two results in the form

x  1.96( / n )    x  1.96( / n ) (2)

Similarly, from the table of normal areas 99% of the area lies between – 2.58 and 2.58. This
is equivalent to the form

x  2.58( / n )    x  2.58( / n ) (3)

Thus we can say that (2) is the 95% confidence interval and (3) is the 99% confidence
interval.

2.5% 2.5% 0.5% 0.5%

95% 99%

-1.96  1.96 -2.58  2.58

From the figure we can say with 95% confidence that if the hypothesis is true, the value of z
for an actual sample lies between – 1.96 to1.96. Since area under the normal curve between
these values is 0.95. However if the value of z for random sample lies outside this range we
can conclude that the probability of the happening of such event is only 0.05 if the given
hypothesis is true.

The total shaded area 0.05 being the level of significance of the test represents the probability
of making type I error (rejecting the hypothesis when it should have been accepted). The set
of values of zoutside the range – 1.96, 1.96 constitute the critical (significant) region (or) the
region of rejecting the hypothesis whereas the values of z within the same range constitute the
insignificant region or the region of acceptance of the hypothesis.

8
One Tailed and Two tailed Tests:

In our test of acceptance or non-acceptance of a hypothesis we concentrated on the value of z

on both sides of the mean. This can be categorically stated that the focus of attention lies in
the two “tails” of the distribution and hence such a test is called a two tailed test.

Sometimes we will be interested in the extreme values to only one side of the mean in which
case the region of significance will be a region to one side of the distribution. Obviously the
area of such a region will be equal to the level of significance itself. Such a test is called a
one tailed test.

The critical value of z in respect of a one tailed test are [-1.645 or 1.645], [-2.33 or 2.33] at
5% and 1% level of significance respectively.

Note:

The following table will be used for working problems

Critical value of z
Test
5% level 1% level
One tailed test -1.645 or 1.645 -2.33 or 2.33
Two tailed test -1.96 and 1.96 -2.58 and 2.58

Examples: A manufacturer claimed that at least 95% of the equipment which he supplied to
a factory conformed to specifications. An examination of a sample of 200 pieces of
equipment revealed that 18 of them were faulty. Test his claim at 1% and 5% level of
significance.

Solution: Let p = Probability of equipment supplied met the specification i.e. p = 0.95 q =
0.05

H0: p = 0.95 i.e., the manufacturer claim is correct

H1: p< 0.95 i.e., the manufacturer claim is not correct.

We choose one tailed test:

n = 200

Expected number of equipment meets the specification, μ= np

μ= 200(0.95) = 190

Observed number of equipment meeting the specification x = 200 – 18 = 182

Test statistic:

x 182  190

z   2.6
npq 200  0.95  0.05

9
 z  2.6

Decision:

(i) At 1% level of confidence z0.01  (2.33or 2.33)

z  2.6  z0.01 (2.33) or z  2.6  z0.01

i.e., z value lies in the critical region, H0 is rejected.

(ii) At 5% level of confidence z0.05  1.645

z  2.6  z0.05 (1.645)

H0 is rejected.

Conclusion: We conclude that the manufacturer claim is not correct at 5% and 1% level of
significance.

Example: It has been found from experience that the mean breaking strength of a particular
brand of thread is 275.6 gms with standard deviation of 39.7 gms. Recently a sample of 36
pieces of thread shows a mean breaking strength of 253.2 gms can one conclude at level of
significance of 5% and 1% that the thread has become inferior.

Solution: H0: μ = 275.6 gms

H1: μ< 275.6 gms

We choose one tailed test

n = 36

Expected mean breaking strength, μ = 275.6

Observed mean breaking strength, x = 253.2

Test statistic:

x 253.2  275.6

z   3.38
/ n 39.7 / 36

Decision:

(i) At 5% level of significance z0.05  1.645

z  3.38  3.38  z0.05

H0 is rejected.

10
(ii) At 1% level of confidence z0.01  2.33

z  3.38  3.38  z0.01

H0 is rejected.

Conclusion: At 5% and 1% level of significance we conclude that the thread has become
inferior.

Example: In an examination given to students at a large number of different schools the

mean grade was 74.5 and standard deviation grade was 8. At one particular school 200
students took the examination the mean grade was 75.9. Discuss the significance of this result
at both 5% and 1% level of significance.

Solution: H0: μ = 74.5, there is no significant difference in the mean grades.

H1: μ  74.5

We choose two tailed test

Here n = 200

Expected mean breaking strength, μ = 74.5

Observed mean breaking strength, x = 75.9

Test statistic:

x 75.9  74.5

z   2.475
/ n 8 / 200

Decision:

(i) At 5% level of significance z0.05  (1.96 and 1.96)

z  2.475  z0.05

H0 is rejected.

(ii) At 1% level of confidence z0.01  (2.58 and 2.58)

z  2.475  z0.001

H0 is accepted.

Conclusion: At 5% level of significance we conclude that there is a significant difference of

the mean grades in the tests but not at 1% level of significance.

11
Test of Significance for difference of means:

Let 1 and 2 be the mean of the populations. Let ( x1 ,  1 ) ; ( x2 ,  2 ) be the mean and S. D. of
two large samples of size n1 and n2 respectively. We wish to test the null hypothesis H0 that
there is no difference between the population means. That is H0: 1 = 2.

( x1  x2 )
The statistic for this test is given by z 
 12 / n1   22 / n2

Also confidence limits for the difference of means of population are

( x1  x2 )  zc  12 / n1   22 / n2

We adopt the same procedure for testing the null hypothesis by using one tailed test or two
tailed test.

Corollary: If the samples are drawn from the same population then  1   2  

( x1  x2 )
z .
 1 / n1  1 / n2

Example: A sample of 100 bulbs produced by a company A showed a mean life of 1190 hrs
and a standard deviation 90 hrs. Also a sample of 75 bulbs produced by a company B showed
a mean life of 1230 hrs and a standard deviation 120 hrs. Is there a difference between mean
life time of bulbs produced by two companies at 5% and 1% level of significance.

Solution: H0: There is no significant difference between mean life time of bulbs produced by
two companies.

Given: x1  1190 ;  1  90 ; n1  100 (Company A)

x2  1230 ;  2  120 ; n2  75 (Company B)

( x2  x1 )
Test statistic: z 
 / n1   22 / n2
2
1

(1230  1190)
z  2.42
902 /100  1202 / 75

We choose two tailed test

Decision: (i) At 5% level of significance z0.05  (1.96 and 1.96)

z  z0.05

H0 is rejected

12
(ii) At 1% level of significance z0.01  (2.58 and 2.58)

z  z0.01

H0 is accepted

Conclusion: At 1% level of significance we conclude that there is no significant difference

between mean life time of bulbs produced by both the companies but not at 5% level of
significance.

Example: The mean of two large samples of 1000 and 2000 members are 168.75 cms and
170 cms respectively, can the samples be regards as drawn from the sample population of
standard deviation 6.25 cms.

Solution: H0: There is no significant difference between mean of two samples i.e., samples
drawn from the same population.

Given: x1  168.75 ; n1  1000

x2  170 ; n2  2000  1   2  6.25

( x2  x1 )
Test statistic: z 
 / n1   22 / n2
2
1

(170  168.75)
z  5.16
6.25 /1000  6.25 / 2000

We choose two tailed test

Decision: (i) At 5% level of significance z0.05  (1.96 and 1.96)

z  z0.05

H0 is rejected

(ii) At 1% level of significance z0.01  (2.58 and 2.58)

z  z0.01

H0is rejected

Conclusion: At 5% and 1% level of significance we conclude that the samples cannot be

regarded as drawn from same population.

13
Example: A random sample for 1000 workers in company has mean wage of Rs 50 per day
and standard deviation of Rs 15. Another sample of 1500 workers from another companyhas
mean wage of Rs 45 per day and standard deviation Rs 20. Does the mean rate of wages
varies between the two companies? Find 95% confidence limits for the difference of the
mean wages of the population of the two companies.

Solution: H0: There is no significant difference in the mean wages between the two
companies.

Given: x1  50 ;  1  15 : n1  1000

x2  45 ;  1  20 n2  1500

( x1  x2 )
Test statistic: z 
 12 / n1   22 / n2

(50  45)
z  7.1306
20 /1500  15 /1000

We choose two tailed test

Decision: (i) At 5% level of significance z0.05  (1.96 and 1.96)

z  z0.05

H0 is rejected

(ii) At 1% level of significance z0.01  (2.58 and 2.58)

z  z0.01

H0 is rejected

Conclusion: At 5% and 1% level of significance we conclude there is a significance

difference in the mean wages between the two companies.

95% confidence limits for the difference of mean wages is given by

( x1  x2 )  1.96  12 / n1   22 / n2

 5  1.96(0.7012)  5  1.374

= 3.626 and 6.374

Thus with 95% confidence the difference of mean wages of population between two
companies lies between Rs 3.62 and Rs 6.32.

14
Example: The mean and standard deviation of the maximum loads supported by 60 cables
are 11.09 tonnes and 0.73 tonnes respectively. Find (a) 95% (b) 91% confidence limits for
mean of the maximum loads of all cables produced by the company

Solution: (a) 95% confidence limit

    0.75 
x  1.96    11.09  1.96    11.09  0.18
 n  60 

Thus 10.91 tonnes to 11.27 tonnes are 95% confidence limits.

(b) 99% confidence limit

    0.75 
x  2.58    11.09  2.58    11.09  0.24
 n  60 

Thus 10.85 tonnes to 11.33 tonnes are 99% confidence limits.

Example: The mean and standard deviation of the diameters of a sample of 250 rivet heads
manufactured by company are 7.2642 mm and 0.0058 mm respectively. Find (a) 99% (98%
(c) 95% (d) 90% confidence limits for the mean diameter of all the rivet heads manufactured
by company.

Solution: Given: x  7.2642 ;   0.0058 : n  250

Given 99% 98% 95% 90%

zc 2.58 2.33 1.96 1.645
  
Confidence limits: x  zc  
 n

(a) 7.2642  2.58  0.00037   7.2642  0.00095

(b) 7.2642  2.33  0.00037   7.2642  0.00086

(c) 7.2642  1.96  0.00037   7.2642  0.00073

(d) 7.2642  1.645  0.00037   7.2642  0.00061

15
Test of Significance for Large Samples:

Test of Significance of Properties:

Let us take N samples each having n members. Let p be the probability of success of each
member and q of failure so that p + q = 1. The frequencies of samples with successes 0, 1, 2,
3, ………n are the terms of the binomial expansion of N (q + p)n. Thus the binomial
distribution is regarded as the sampling distribution of the number of successes in the sample.

We know that  = np and S. D. = npq .

Let us consider the proportion of successes

np
(i) mean proportion of successes =  p
n

npq pq
(ii) S. D. or S. E. proportion of successes = 
n n

Test statics:

(1) In proportion of successes

x p
z
pq n

Where x: Observed proportion of success

p = Expected proportion of success

n: Sample size

(2) In terms of number of success

x x  np
z 
 npq

Where x: Observed proportion of success

μ = np = Expected number of success

  S .D.  npq

16
Probable limits:

(1) At 1% level of significance probable limits are given by

pq
p  2.58
n

Where p: probability of success

pq
: S. E. proportion of successes
n

(2) At 5% level of significance probable limits are given by

pq
p  1.96
n

Example: A coin is tossed 1000 times and head turns up 540 times. Decide an hypothesis
that the coin is unbiased.

Solution: H0: Coin is unbiased

Given:

p = probability of success (getting head) = 1/2

q = 1/2

n = 1000

Observed number of success = x = 540

Expected number of success = μ = 1000 × 1/2 = 500

x 540  500

Test static: z    2.53
npq 1000  1/ 2 1/ 2

Decision: At 1% level of significance z0.01  (2.58 and 2.58)

z  2.53  z0.01

Therefore H0 is accepted.

Conclusion: Coin is unbiased.

17
Example: Result extracts revealed that in a certain school over a period of five years 725
students had passed and 615 students had failed. Test the hypothesis that success and failure
are in proportion.

Solution: H0: Success and failure are in proportion

p = probability of success (passed) = 1/2

q = 1/2

n = 725 + 615 = 1340 (total number of students)

725
Observed proportion of success, x   0.54
1340

Expected proportion of success = p = 1/2 = 0.5

x p 0.54  0.5
Test static: z    3.005
pq 0.5  0.5
n 1340

Decision: At 1% level of significance z0.01  (2.58 and 2.58)

z  3.001  z0.01

Therefore H0 is rejected.

Conclusion: Success and failure are not in equal proportion.

Example: In a sample of 500 men it was found that 60% of them had over weight. When we
can about the proportion of people having over weight in the population.

Solution: Proportion / probable percentage limits of people having over weight to be found

Probable limits at 1% level of significance is given by

pq
p  2.58
n

p = probability of persons having over weight = 60/100 = 0.6

q = 0.4

pq 0.6  0.4
p  2.58  0.6  2.58  0.6  0.0565
n 500

The probable percentage limits of people having over weight is 54.35% to65.65%.

18
Example: A survey was conducted in a slum locality of 2000 families by selecting a sample
of size 800. It was revealed that 180 families were illiterate. Find the probable limits of
illiterate families in the population of 2000.

Solution: Probable limits at 1% level of significance is given by

pq
p  2.58
n

p = probability of illiterate families = 180/800 = 0.225

q = 0.775

pq 0.775  0.225
p  2.58  0.225  2.58
n 800

= 0.187 and 0.263

The probable limits of illiterate families in the population o 2000 is = 0.187 × 2000 to 0.263×
2000

i.e., 374 to 526 are probable illiterate families.

Example: In 324 throws of a six faced die an odd number turned up 181 times. It is
reasonable to think that die is an unbiased one.

Solution: H0: The die is an unbiased one.

Observed proportion of success (odd number) = x = 181/324 = 0.56

Expected proportion of success, p = 3/6 = 1/2 = 0.5

q = 1/2

n = 324

x p 0.56  0.5
Test static: z    2.22
pq 0.5  0.5
n 324

Decision: At 1% level of significance z0.01  (2.58 and 2.58)

z  2.22  z0.01

Therefore H0 is accepted.

Conclusion: The die is unbiased.

19
Example: A die is thrown 9000 times and a throw of 3 or 4 was observed 3240 times. Show
that the die cannot be regarded as an unbiased one.

Solution: H0: The die is an unbiased one.

Observed proportion of success (odd number) = x = 3240/9000 = 0.36

Expected proportion of success, p = 2/6 = 1/3 = 0.33

q = 2/3n = 9000

x p 0.36  0.33
Test static: z    6.03
pq 0.33  0.67
n 9000

Decision: At 1% level of significance z0.01  (2.58 and 2.58)

z  6.03  z0.01

Therefore H0 is rejected.

Conclusion: Thus the die cannot be regarded as an unbiased one.

Example: A sample of 100 days is taken from meteorological records of a certain district and
10 of them are found to be foggy. What are the probable limits of the percentage of foggy
days in the district?

Solution: Probable limits are given by

pq
p  2.58
n

p = probability of foggy days = 10/100 = 0.1

q = 0.9

pq 0.1  0.9
p  2.58  0.1  2.58  0.1  0.0774
n 100

= 0.0226 and 0.1774

Percentage of foggy days lies between 2.26% and 17.74%.

20
Example: A random sample of 500 apples was taken from a large consignment and 65 were
found to be bad. Estimate the proportion of bad apples in the consignment as well as the
standard error of the estimate. Also find the percentage of bad apples in the consignment.

Solution: Probable limits are given by

pq
p  2.58
n

p = probability of bad apples = 65/500 = 0.13

q = 0.87

pq 0.13  0.87
p  2.58  0.13  2.58
n 500

= 0.0913 and 0.1687

Percentage of bad apples in the consignment lies between 9.13% and 16.87%.

Example: In a locality of 18000 families a sample of 840 families was selected at random.
Of these 840 families, 206 families were found to have monthly income of Rs 2500 or less. It
was desired to estimate how many of the 18000 families have monthly income of Rs 2500 or
less.

Solution: p = 206/840 = 0.245, q = 0.755 n = 840

pq
99% confidence level is given by p  2.58
n

pq 0.245  0.755
p  2.58  0.245  2.58
n 800

= 0.2063 and 0.2837

= 20.63% and 28.37%

Probable limits of 18000 families or given by 0.2063 × 18000 and 0.2837 × 18000 that is
3713 to 5107 families are likely to have monthly income of Rs 2500or less.

Example:In a hospital 230 females and 270 males were born in a year. Do these figures
confirm the hypothesis that sexes are born in equal proportion?

Solution: H0: The female and male babes are born in equal proportion.

230
Observed proportion of success (female born), x   0.46
500

Expected proportion of success = p = 1/2 = 0.5

21
q = 1/2 n = 230 + 270 = 500

x p 0.5  0.46
Test static: z    1.786
pq 0.5  0.5
n 500

Decision: At 1% level of significance z0.01  (2.58 and 2.58)

z  1.786  z0.01

Therefore H0 is accepted.

Conclusion: The female and male babes born in equal proportion.

Example: An unbiased coin is thrown n times. It is desired that the relative frequency of
appearances of heads should be lies between 0.49 and 0.51. Find the smallest value of n that
will ensure the result with (a) 95% confidence and (b) 90% confidence.

Solution: (a) 95% confidence level

pq 0.5  0.5  1 
p  1.96  0.5  1.96  0.5  1.96   = 0.51 or 0.49
n n 2 n 

 1   1 
0.5  1.96    0.51 and 0.5  1.96    0.49
2 n 2 n

 1  1.96
1.96    0.01 or n   98
2 n 0.02

Therefore n = 9604

(b) 90% confidence level

pq 0.5  0.5  1 
p  1.645  0.5  1.645  0.5  1.645   = 0.51 or 0.49
n n 2 n 

 1   1 
0.5  1.645    0.51 and 0.5  1.645    0.49
2 n 2 n

 1  1.645
1.645    0.01 or n   82.25
2 n 0.02

Therefore n = 6765.0625 ≈ 6765.

22
Test of Significance for Difference of Properties (Attribute) for two Samples:

Let p1 and p2 be the sample proportions in respect of an attribute corresponding to two large
samples of size n1 and n2 drawn from two populations. We wish to test the null hypothesis
H0that there is no difference between the population with regard to the attribute.

The statistic for this test is given by

( p1  p2 )
z
pq 1 / n1  1 / n2 

n1 p1  n2 p2
where p  and q  1  p .
n1  n2

We adopt the same procedure for testing the null hypothesis by using one tailed test or two
tailed test.

Example: In an exit poll enquiry it was revealed that 600 voters in one locality and 400
voters from another locality favoured 55% and 48% respectively a particular party to come to
power. Test the hypothesis that there is a difference in the localities in respect of the opinion.

Solution: H0: There is no difference in opinion in the locality.

55 48
Given: p1   0.55 ; p2   0.48 ; n1  600 ; n2  400
100 100

n1 p1  n2 p2 (600  0.55)  (400  0.48)

Therefore p    0.522
n1  n2 600  400

q = 0.478

We choose two tailed test

Rest statistic:

( p1  p2 ) 0.55  0.48
z   2.171
pq 1/ n1  1/ n2  (0.522  0.478)(1 600  1 400)

Decision:

(1) At 1% level of significance z0.01  (2.58 and 2.58)

z  2.171  z0.01

Therefore H0 is accepted

23
(2) At 5% level of significance z0.05  (1.96and 1.96)

z  2.171  z0.05

Therefore H0 is rejected

Conclusion: At 1% level of significance we conclude that there is no difference in opinion in

the locality but not at 5% level of significance.

Example: One type of aircraft is found to develop engine trouble in 5 flights out of a total of
100 and another type in 7 flights out a total of 200 flights. Is there a significance difference in
two types of aircrafts so as engine defects are concerned?

Solution: H0: There is no difference in opinion in two types of aircrafts as engine defects are
concerned.

5
Given: p1 (proportion of defects I type)   0.05 ;
100

7
p2 (proportion of defects II type)   0.035 ; n1  100 ; n2  200
200

n1 p1  n2 p2 (100  0.05)  (200  0.035)

Therefore p    0.04
n1  n2 100  200

q = 0.96

We choose two tailed test

( p1  p2 ) 0.05  0.035
Test statistic: z    0.625
pq 1/ n1  1/ n2  (0.04  0.96)(1 100  1 200)

Decision:

(1) At 1% level of significance z0.01  (2.58 and 2.58)

z  0.625  z0.01

Therefore H0 is accepted

(2) At 5% level of significance z0.05  (1.96and 1.96)

z  0.625  z0.05

Therefore H0 is accepted

Conclusion: At 1% and 5% level of significance we conclude that there is no significance

difference in two types of aircrafts as engine defects are concerned.
24
Example: Random sample of 1000 engineering students from a city A and 800 from city B
were taken. It was found that 400 students in each of the sample were from payment quota.
Does the data reveal a significant difference between the two cities in respect of payment
quota students?

Solution: H0: There is no significant difference between two cities in respect of payment
quota students.

400 400
Given: p1   0.4 ; p2   0.5 ; n1  1000 ; n2  800
1000 800

n1 p1  n2 p2 (1000  0.4)  (800  0.5) 4

Therefore p   
n1  n2 1000  800 9

q = 5/9

We choose two tailed test

Rest statistic:

( p1  p2 ) 0.4  0.5
z   4.243
pq 1/ n1  1/ n2   4 5
   (1 1000  1 800)
9 9

Decision:

(1) At 1% level of significance z0.01  (2.58 and 2.58)

z  4.243  z0.01

Therefore H0 is rejected

(2) At 5% level of significance z0.05  (1.96and 1.96)

z  4.243  z0.05

Therefore H0 is rejected

Conclusion: At 1% and 5% level of significance we conclude that there is no significant

difference between two cities in respect of payment quota students.

25
Test of Significance for Small samples:

In case of large samples, sampling distribution follows a normal distribution which is not true
in the case of small samples. We introduce the concept of degree of freedom for discussing
student’s t distribution.

Degree of freedom:

The number of degree of freedom (d.f.) usually denoted by v is the number of values in a set
which may be assigned arbitrarily. It can be interpreted as the number of independent values
generated by a sample of small size for estimating a population parameter.

Example:

Let us suppose that we need to find 3 numbers whose sum is 25 i.e., to find a, b, c such that a
+ b + c = 25. We can arbitrarily assign values to any two of the variables a, b, c and hence
these are the degree of freedom. That is to say that d.f(v) = 2. If there are n observations d.f. is
equal to n – 1.

Suppose that we are finding the mean of a sample of size n comprising values x1, x2, x3,
…..xn. We use all the n values to compute the sample mean x and x is said to have n degree
of freedom. Suppose we are finding the sample variance we use the n values

( x1  x ) 2 , ( x2  x ) 2 ,……….. ( xn  x ) 2 .

But these values do not have n degree of freedom as they are depend on a fixed value x
which has already been computed. Hence the sample variance is said to have (n – 1) degree
of freedom. If we compute another statistic based on the sample mean and variance, that
statistic is said to have (n - 2) degree of freedom and so on.

In general the number of degree of freedom v = n – k where n is the number of observations

in the sample and k is the number of constraints / number of values which are pre determined.

Student’s t distribution:

Let xi (i  1, 2,3....n) be a random sample of size n drawn from a normal population with
mean  and variance  2 . The statistic t is defined as follows

x  x 
t  n
s/ n s

1 n
Here x   xi is the sample mean.
n i 1

1 n
s2   ( xi  x ) is the sample variance.
2

n  1 i 1

v  n  1 denote the number of degree of freedom of t.

26
The statistic t follows the student’s t distribution with (n – 1) degree of freedom having the
probability density function

y0
y  f (t )  ( v 1)/2
 t2 
1  
 v

where y0 is a constant such that the area under the curve is unity.

Note(1):

x 
Statistict is also defined as follows t  n 1


Note(2):

(v  1)

The constant y0 present in p.d.f. is given by y0  2 so that the p.d.f. of student’s t
 v  v / 2
distribution is given by

(v  1)
  ( v 1)/ 2
2  t2 
y  f (t )  1   ;   t   .
 v  v / 2   v

Note(2):

If v is large ( v  30 ) the graph of f(t) closely approximates standard normal curve. In other
words we can say that t is normally distributed for large samples.

Test of significance of difference between sample means:

Consider two independent samples xi and y j (i  1, 2, 3,.....n1 ; j  1, 2,3,......n2 ) drawn from a

normal population.

Let  x ,  x  and  y ,  y  respectively be mean and variance of two samples.

Let  and  be population mean and variance, then

xy
t
1 1
S 
n1 n2

1  n1 n2
2
  xi  x     y j  y  
2
where S 2

n1  n2  1  i 1 j 1 

27
and  (degree of freedom)  n1  n2  2

n1 S12  n2 S22
(or) S 
2

n1  n2  2

Confidence limits for the population mean μ

S
95% confidence limits for μ is x  t0.05
n

S
99% confidence limits for μ is x  t0.01
n

Working Procedure:

Case I: Students t test for a sample of means

Here we test the hypothesis whether sample mean ( x ) differs significantly from population
mean (μ).

x 
Step 1: Compute t  n and consider t
S

Step 2:(a) If t > t0.05 the difference between x and μ is said to be significant at 5% level of
significance.

(b) If t > t0.01 the difference between x and μ is said to be significant at 1% level of
significance.

Case II: Students t test for difference between sample means

Here we test the hypothesis whether difference between the sample means is significant.

xy
Step 1: Compute t  and consider t
S 1 n1  1 n2

Step 2:(a) If t > t0.05 the difference between the sample means is said to be significant at 5%
level of significance.

(b) If t > t0.01 the difference between the sample meansis said to be significant at 1% level of
significance.

Note: If t is less than the table value at a certain level of significance, the data is said to be
consistent with the hypothesis.

28
Example: A machine is expected to produce nails of length3 inches. A random sample of 25
nails gave an average length of 3.1 inch with standard deviation 0.3. Can it be said that the
machine is producing nails as per specification? ( t0.05 for 24 d.f. is 2.064).

Solution: H0: There is no difference between population mean (average length of nails) and
sample mean (i.e. the machine is producing nails as per the specification)

Level of significance: At 5% of level of significance t0.05 =2.064 for 24 d.f.

Given:   3, x  3.1, n  25,  0.3

Test static:

x  3.1  3
t  25  1.67  2.064
s/ n 0.3

Conclusion: Hypothesis that the machines is producing nails as per specification is accepted
at 5% level of significance.

Example: Ten individuals are chosen at random from a population and their heights in inches
are found to be 63, 63, 66, 67, 68, 69, 70, 70, 71, 71. Test the hypothesis that the mean
heights of the universe is 66 inches.( t0.05 =2.262 for 9 d.f. ).

Solution: H0: There is no significant difference between population mean height and sample
mean height (i.e. the mean height of the universe is 66 inches)

Level of significance: At 5% of level of significance t0.05 =2.262 for 9d.f.

Given:

n  10,   66, x 
 x  63  63  66  67  68  69  70  70  71  71  678  67.8,
n 10 10

1
s2 
n 1
 ( x  x )2
1 (63  68.7)  (63  68.7)  (66  68.7)  (69  68.7)  (68  68.7) 
2 2 2 2 2

    9.067
9 (69  68.7)2  (70  68.7) 2  (70  68.7) 2  (71  68.7) 2  (71  68.7) 2 

s  3.04

Test static:

x  67.8  66
t  10  1.89  2.262  t0.05
s/ n 3.011

Conclusion: Hypothesis is accepted at 5% level of significance.

29
Example: A sample of 10 measurements of the diameter of a space gave a mean of 12 cm
and a standard deviation 0.15 cm. Find the 95% confidence limit for the actual diameter.

Solution: Given: n = 10, x  12 ,s = 0.15

Also t0.05 =2.262 for 9 d.f.

s 0.15
Confidence limits x  t0.05  12  (2.262)  12  0.1073
10 10

Thus 11.893 to 12.107 is the confidence limit for the actual diameter.

Example: A certain stimulus administered to each of the 12 patients resulted in the following
change in blood pressure 5, 2, 8, -1, 3, 0, 6, -2, 1, 5, 0, 4. Can it be concluded that the
stimulus will increase the blood pressure. ( t0.05 for 11 d.f. is 2.201).

Solution: H0: There is no significance difference between population mean and sample mean
(i.e. no change in blood pressure)

H1: Stimulus will increase the blood pressure.

(One tailed test)

x
 x  2.5833
n

1
s2 
n 1
 ( x  x )2  9.538
s  3.088

x 
Test static: t  n
s

(   0 , assuming the stimulus administered is not accompanied with increase in blood

pressure)

2.5833  0
t 12  2.8979  2.9  2.201
3.088

H0 is rejected at 5% level of confidence

The stimulus in general is accompanied with increase in blood pressure.

30
Example: A group of boys and girls were given an intelligence test. The mean score,
standard deviation score and numbers in each group are as follows

Boys Girls
Mean 74 70
S.D. 8 10
n 12 10
Is the difference between the means of the two groups significant at 5% level of significance?
( t0.05 for 20 d.f. is 2.086).

Solution: H0: There is no significant difference between the means of two groups.

Given: x  74, s1  8, n1  12 (Boys)

y  70, s2  10, n2  10 (Girls)

xy
Test static: t 
s 1/ n1  1/ n2

n1 s12  n2 s22 1768

where s 2    88.4
n1  n2  2 20

Therefore s = 9.4

74  70
t  0.994  t0.05  2.086
9.4 1/12  1/10

Decision: H0 is accepted at 5% level of significance.

Conclusion:

Thus there is no significance difference between the means of the two groups.

Example: Two types of batteries are tested for their length of life and the following results
were obtained

Battery A: n1 = 10, x1  500 ,  12  100

Battery B: n2 = 10, x2  500 ,  22  121

Compute student’s t and test whether there is a significant difference in two means.

Solution: H0: There is no significant difference between the mean length of life of Battery A
and Battery B.

Given: n1  10, x  560,  12  100, (Battery A)

n2  10, x2  500,  22  121, (Battery B)

31
 n1 s12  n2 s22
s2   122.73
n1  n2  2

Therefore s = 11.0805

xy
Test static: t   12.11  all values in t - table
s 1/ n1  1/ n2

H0 is rejected

There is significant difference between the mean length of life of Battery A and Battery B.

Example: A group of 10 boys fed on a diet A and another group of 8 boys fed on a different
diet B for a period of 6 months recovered the following increase in weight (lbs)

Diet A 5 6 8 1 12 4 3 9 6 10
Diet B 2 3 6 8 10 1 2 8
Test whether diet A and B differ significantly regarding their effects on increase in weight.

( t0.05 for 16 d.f. is 2.12).

Solution: H0: There is no significant difference between diet A and diet B on increase in
weight.

Let x correspond to diet A and y correspond to diet B.

x
 x  64  6.4 ; y
 y  40  5
n1 10 n2 8

s 2

 (x  x )   ( y  y)
2 2

 11.525
n1  n2  2

s  3.395

xy
Test static: t   0.87  t0.05  2.12
s 1/ n1  1/ n2

Decision: H0 is accepted.

Conclusion: Two diets do not differ significantly regarding their effects on increase in
weight.

32
Example: Two horses A and B were tested according to the time(in seconds) to run a
particular race with the following results

Horse A 28 30 32 33 33 29 34
Horse B 29 30 30 24 27 29
Test whether you can discriminate between two horses.

Solution: H0: There is no significant difference between two horses.

Let x correspond horse A and y correspond to horse B.

x
 x  219  31.3 ; y
 y  169  28.2
n1 7 n2 6

s2 
 (x  x )   ( y  y )
2 2

 5.2973  s  2.3011
n1  n2  2

xy
Test static: t   2.42  t0.05  2.12
s 1/ n1  1/ n2

Decision: H0 is rejected.

Conclusion: There is significant difference between two horses.

Example: Eleven school boys were given a test in drawing. They were given a month further
tuition and a second test of equal difficulty was held at the end of it. Do the marks give
evidence that the students have been benefitted by extra coaching? (Given t0.05 for 10 d.f. is
2.228)

Boys 1 2 3 4 5 6 7 8 9 10 11
Marks(I test) 23 20 19 21 18 20 18 17 23 16 19
Marks(II test) 24 19 22 18 20 22 20 20 23 20 17

Solution: Let x denote the difference in marks in the two test of eleven students

x 1 -1 3 -3 2 2 2 3 0 4 -2

x
 x  11  1
n 11

1 0  4  4  16  1  1  1  4  1  9  9 50
s2 
n 1
 ( x  x )2 
10

10
 5  s  2.236

H0: There is no significant difference in Marks in the two tests (i.e. Students are not
benefitted by extra teaching)

33
 x  1 0
Test Static: t  n 11  1.483 < t0.05 = 2.228
s 2.236

(Taking μ = 0, let us suppose that the students are not benefitted by extra coaching)

Decision: H0 is accepted.

Conclusion: Students are not benefitted by the extra coaching.

Example: A random sample of 10 boys had the following IQ 70, 120, 110, 101, 88, 83, 95,
98, 107, 100

Do these data support the assumption of a population mean IQ of 100at 5% level of

significance. (Given t0.05 for 9 d.f. is 2.262)

Solution: H0: There is no significant difference between population mean IQ and sample IQ
(i.e. population mean IQ is 100)

Given: n = 10, x  97.2 , s  14.2735 ,   100

x 
Test static: t  n  0.62
s

t  0.62  2.262

Decision: H0 is accepted.

Conclusion: The population mean IQ is 100.

Example: A sample of 12 measurements of the diameter of metal ball gave the mean 7.38
mm with standard deviation 1.24 mm. Find (i) 95% (ii) 99% confidence limits for the actual
diameter.

(Given t0.05 for 11 d.f. is 2.20 and t0.01 for 11 d.f. is 3.11)

Solution: (i) 95% confidence limits for the actual diameter is given by

 s   1.24 
x   t0.05  7.38    (2.2)  7.38  0.7875
 n  12 

Thus 6.5925 mm to 8.1675 mm are the 95% confidence limits for the actual diameter.

(ii) 99% confidence limits for the actual diameter is given by

 s   1.24 
x   t0.01  7.38    (3.11)  7.38  1.1132
 n  12 

Thus 6.2668 mm to 8.4932 mm are the 99% confidence limits for the actual diameter.

34
Chi – square (  2 ) distribution:

 2 distribution provides a measure of correspondence between the theoretical and observed

frequencies.

If Oi (i = 1, 2, 3, ……..n) and Ei respectively denotes a set of observed and estimated

frequencies the quantity chi – square denoted by  2 is defined as follows

n
(Oi  Ei ) 2
2   , degree of freedom = n - 1
i 1 Ei

Chi – square test as a test of goodness of fit:

If the calculated value of  2 is less than the table value of  2 at a specified level of
significance the hypothesis is accepted, otherwise the hypothesis is rejected.

Example: A die is thrown 264 times and the number appearing on the face (x) follows the
following frequency distribution

x 1 2 3 4 5 6
f 40 32 28 58 54 60
Calculate the value of  2 .

Solution: Given values are observed frequencies (Oi)

Assuming that die is unbiased, expected number of frequencies for numbers 1, 2, 3, 4, 5, 6

264
are  44 each
6

Number on die 1 2 3 4 5 6
Oi 40 32 28 58 54 60
Ei 44 44 44 44 44 44

(O  Ei ) 2  40  44    32  44   28  44    58  44    54  44    60  44 
2 2 2 2 2 2
n
  i
2

i 1 Ei 44

968
2   22
44

35
Example: Five dice were thrown 96 times and the number 1, 2 or 3 appearing on the face of
the dice follows the frequency distribution as follows

No. of dice showing 1, 2 or 3 5 4 3 2 1 0

Frequency 7 19 35 24 8 3
Test the hypothesis that the data follows a binomial distribution (or fit a binomial distribution
for the data and test the goodness of fit  2 0.05  11.07 for 5 d.f.)

3 1 1
Solution: Probability of getting 1, 2,or 3 is p   , q
6 2 2

Probability density function is given by

x 5 x 5
n x 1 1 1 1 5
n x
Cx p q  Cx    
5
 Cx   
5
Cx
2 2 2 32

The theoretical frequencies are as follows

1 5
F ( X  0)  96  C0  3
32

1 5
F ( X  1)  96  C1  15
32

1 5
F ( X  2)  96  C2  30
32

1 5
F ( X  3)  96  C2  30
32

1 5
F ( X  4)  96  C4  15
32

1 5
F ( X  5)  96  C5  3
32

Oi 7 19 35 24 8 3
Ei 3 15 30 30 15 3

n
(Oi  Ei )2 16 16 25 36 49 0
2          11.7
i 1 Ei 3 15 30 30 15 3

 2  11.70   0.05
2
 11.07

Hypothesis is rejected

i.e. Fitness is not good.

36
Example: Fit Poison distribution for the following data and test the goodness of fit given that
 2 0.05  7.815 for 3 d.f.

x 0 1 2 3 4
f 122 60 15 2 1
e m m x
Solution: Probability distribution function of poison distribution is where m is mean
x!

x
fxi i

0  60  30  6  4
 0.5
f i 200

e0.5 (0.5) x (0.6065)(0.5) x

Poison distribution reduces to 
x! x!

Theoretical frequencies are given by

(0.6065)(0.5) 0
F ( X  0)  200   121.3  121
0!

(0.6065)(0.5)1
F ( X  1)  200   60.65  61
1!

(0.6065)(0.5) 2
F ( X  2)  200   15.16  15
2!

(0.6065)(0.5)3
F ( X  3)  200   2.53  3
3!

(0.6065)(0.5) 4
F ( X  4)  200   0.31  0
4!

x 0 1 2 3 4
Oi 122 60 15 2 1
Ei 121 61 15 3 0

Combining previous frequency if Ei = 0

Oi 122 60 15 3
Ei 121 61 15 3

 2  0.025   0.05
2
 7.815

Therefore H0 is accepted.

37
Example: A sample analysis of examination results of 500 students was made. It was found
that 220 students had failed, 170 students had secured third class, 90 students had secured
second class and 20 students secured first class. Do these figures support the general
examination result, which is in ration 4:3:2:1 for respective categories? (  2 0.05  7.81 for 3
d.f.)

Solution: H0: The figures support the general examination result.

Observed frequencies are 220, 170, 90, 20.

Expected values are

4
No of failure  500  200
10

3
No of third class  500  150
10

2
No of second class  500  100
10

1
No of first class  500  50
10

Oi 220 170 90 20
Ei 200 150 100 50

 220  200  170  150   90  100   20  50 

2 2 2 2

 
2
  
200 150 100 50

 2  23.67   0.05
2
 7.81

Therefore H0 is rejected.

38