Sie sind auf Seite 1von 30

78

CHAPTER - 4
Sampling Distributions and Tests of Hypothesis
4.1 Concepts of Population and Samples:
4.1.1 Population
A group of individuals under investigation is called a Population.
In Statistical analysis, the word population refers to general magnitude and the study of
variation with respect to one or more characteristics relating to the individuals in a
group is known as population. Thus, the population is nothing but an aggregate study of
group of objects, living or non-living, at a particular time period of the given
geographical location.

The population is further classified into two types: namely, Finite population, which
constitutes a finite or countable finite number of objects, for example: the no. of
engineering colleges under ANU, the no. of sugar cane industries in AP state, etc., On the
other hand an infinite population which consists of an infinite no. of objects, for example:
the no of independent trials of an unbiased coin, etc.,

4.1.2 Sample and Sample Size
A finite subset of observations chosen from a given population is known as a sample and
the number of observations in the sample is known as sample size.
- Classification of Samples: Samples are classified into two ways.
- Large Sample: If the size of the sample 30 the sample is said to be large sample.
- Small Sample: If the size of the sample < 30 the sample is said to be small
sample.

4.1.3 Methods of drawing samples
There are two different methods of selecting sample from the given population.
a) Lottery Method: Under this method, all the members (observations) of the
given population are numbered serially from 1, 2,.., N on an uniform slips and
79

put into a box. A required number of members (observations) are selected from
these thoroughly shuffled slips, is known as a sample.
This method is so simple and easy to apply and understand, if and only if the
population size is small. When the population size is large or sufficiently large
this method is not applicable. Moreover this method is not reliable as since
different persons may get different samples from same set of population and
sample, which we are selected, may not be a true representation for the
population characteristics.
b) Random Numbers Method: This is most important and widely applied method
to select a sample from the given population (finite or infinite). The method
consists of the following steps:
1 Let us consider a population of size N observations, say 1, 2, 3, ...., N.
2 The population size N may be a single digit (0,1, .., 9) or a two digit
(00,01,, 99) or a three digit number etc.,
3 A required sample of size n is selected from the well defined random
numbers tables given by Fisher and Yates tables, by neglecting the starting
and the ending number in the population, by moving in any direction, i.e.,
horizontally or vertically down words or diagonally.
4 The sample by this method is a true representative of the population
characteristics.

4.1.4 Random Sample (Finite Population)
Let us consider (x
1
, x
2
, .. , x
n
) be a sample of n observations from a finite population
and is said to be a random sample, if it satisfies the following two conditions:
1 All the sample observations are independent
2 All the sample observations are known chance of being selected in the
sample (i.e.,
N
n
)

4.1.5 Random Sample (Infinite Population)
Let us consider (x
1
, x
2
, .. , x
n
) be a sample of n observations from an infinite
population and is said to be a random sample, if it satisfies the following two conditions:
80

1 All the sample observations are independent
2 All the sample observations are satisfies the probability function associated
with the given population.

4.2 Notation and Terminology
Population mean
Population variance o
2

Population s.d. o
Sample mean
x

Sample variance s
2

Sample s.d. s

4.3 Sampling
This is one of the methods of studying the population and this is most commonly applied
to the data. Under this method, the population is studied on the basis of the
information available in the sample, which is selected from the given population.

4.4 Parameter and Statistic
Any statistical constant of the Population observations is called the Parameter. In other
words, Parameter is a single constant value, which represents the entire population
characteristics. Whereas any statistical constant of the Sample observations is called a
Statistic. In other words, Statistic is a mathematical constant value, which represents
the sample of observations. Statistic is known as a parameter estimate.

4.5 Sampling Distribution
A series of statistic values are arranged in the form of a frequency distribution is known
as a sampling distribution.

More precisely, a sample of size n is selected from a given population of size N. Then
the number of possible samples will be
n
N
C
= k, say. Now calculate a statistic such
as mean, median, mode, standard deviation, etc., on each of the k samples, then we will
81

get a series of k statistic values. These statistic values are arrange in the form of a
frequency table is known as sampling distribution.

Suppose if we calculate the arithmetic mean on k samples then we will get sampling
distribution of sample mean, if we calculate the s. d., on k samples then we will get
sampling distribution of sample s. d., etc.,

4.6 Standard Error
The standard deviation of sampling distribution is known as standard error.

4.7 The Sampling Distribution of the Mean (o Known)
If we want to draw a sample of size n (without replacement) from a finite population of
size N, the total no. of possible samples are C(N,n) =

= k(say)

Example 4.1
A population consists of 4 observations 10, 20, 30, 40 write all the possible samples of
size 2 and then construct the sampling distribution (i) about mean (ii) about variance (iii)
also show that the mean of sample means is equal to the population mean

Solution
Given population observations are 10, 20, 30, 40
A sample of size 2 can be drawn in C(4,2) = 6 ways
Sample distribution about mean and Varience:
S.No. Sample observations Sample mean Sample Variance
1) 10, 20
1
= 15
S
1
2
= 25
2) 10, 30
2
= 20

S
2
2
= 100
3) 10,40
3
= 25 S
3
2
= 225
4)
20, 30
4
= 25 S
4
2
= 25
5) 20, 40
5
= 30 S
5
2
= 100
6) 30, 40
6
= 35 S
6
2
= 35
x
x
x
x
x
x
82

(iii) Mean of Sample means = (15+20+25+25+30+35)/6 = 25= x

Population mean = 10+20+30+40 = 25

4.8 The Central Limit Theorem
If repeated samples of size n are drawn from any infinite population with mean and
variance o
2
, and n is large (n > 30), the distribution of , the sample mean, is
approximately normal, with mean (i.e. ( ) E x = ) and variance o
2
/n (i.e.
2
( ) V x
n
o
= ),
and this approximation becomes better as n becomes larger.

Note
As in the previous illustrating example, we can see the following modifications:
(i) If the population is finite,
2
( ) (1 )
n
V x
N n
o
= ; where (1-n/N) is known as
the finite population correction factor. When N is very big, the factor is
equal to 1.
(ii) If n is small, say less than 30, the sampling distribution is not so normal. A
t-distribution will be used (discussed later).
In the above example, N=4, n=2,
2
( ) (1 )
n
V x
N n
o
= =(1-2/4)(1.6667/2) = 0.4167. If the
population is big (or the sample is drawn with replacement), then
2
( ) V x
n
o
=
=1.6667/2=0.8333.
In this course we assume a big population or sampling with replacement.

Example 4.2
An electrical firm manufactures light bulbs that have a length of life that is
approximately normal distributed with mean equal to 800 hours and a standard
deviation of 40 hours. Find the probability that a random sample of 16 bulbs will have
an average life of less than 775 hours.

x
83

Solution
Let X be the average life of the 16 bulbs. X ~ N( = =
x
800, = =
n
x
2
2
o
o
16
40
2
)
|
|
|
|
.
|

\
|

< =
|
|
.
|

\
|
<

= <
16
40
800 775 775
) 775 ( Z P
X
P X P
x
x
x
x
o

o

00621 . 0 ) 5 . 2 ( = < = Z P

Example 4.3
The mean IQ scores of all students attending a college is 110 with a standard deviation
of 10. (a) If the IQ scores are normally distribution, what is the probability that the
score of any one student is greater than 112? (b) What is the probability that the
mean score in a random sample of 36 students is greater than 112? (c) What is the
probability that the mean score in a random sample of 100 students is greater than 112?

Solution
(a) Let X be the student's IQ score. X ~ N(110,
2
10 )
4207 . 0 ) 2 . 0 (
10
110 112
) 112 ( = > =
|
.
|

\
|
> = > Z P Z P X P

(b) Let
1
X be the mean score of a sample of 36 students.

1
X ~ N( , 110 = =
x

36
10
2 2
2
= =
n
x
o
o )
1151 . 0 ) 2 . 1 (
36
10
110 112
) 112 (
1
= > =
|
|
|
|
.
|

\
|

> = > Z P Z P X P
(c) Let
2
X be the mean score of a sample of 100 students.

2
X ~ N( , 110 =
100
10
2 2
=
n
o
)
84

228 . 0 . 0 ) 2 (
100
10
110 112
) 112 (
2
= > =
|
|
|
|
.
|

\
|

> = > Z P Z P X P
4.9 Theory of Estimation:
4.9.1 Estimation
The process of estimating the unknown values of the population parameters on the
basis of information contained in the sample is known as estimation.

4.9.2 Estimator and Estimate
Let us consider (x
1
, x
2
, .. , x
n
) be a random sample of size n is chosen to from a
population, for which the parameter is an unknown value.
Any statistic or mathematical function that is defined on this sample is known as an
estimator and the particular numerical value, which is assigned to the estimator, is
known as an estimate.
The followings are some examples
Estimator Population parameter
1.
x

2. s
2
o
2

There are two important properties for an estimator, namely, unbiasedness and
efficiency.
An estimator, for example, x , is unbiased if and only if ( ) E x = .
Efficiency: The efficiency of an estimator, for example, x , is given by ( ) V x . The smaller
the ( ) V x , the more accurate will be the x as an estimator.

4.9.3 Point Estimation and Interval Estimation
Let us consider u

be an estimator to the parameter u of the given population.

A point estimate is a single-value estimate of a population parameter, for example
^ ^
; x P p = = . In other words, if the estimator is a single value in the entire real number
scale then such estimation is known as point estimation.

85

An interval estimate of a population parameter gives an interval that may contain the
true value of the parameter with a certain probability (i.e. confidence); for example,
Pr( ) 0.99. a b < < = In other words, if the estimator provides an interval of values
between which the true values of the population parameter will exists is known as
Interval Estimation. The interval of values is known as Confidence Interval and the limit
is a confidence limits.

For a point estimate, both the accuracy and reliability of the estimation are unknown.
For an interval estimate, the width of the interval gives the accuracy and the probability
gives the reliability of the estimation.

Example 4.4
(a) The mean and standard deviation for the quality point averages of a random
sample of 36 college seniors are calculated to be 2.6 and 0.3, respectively. Find a
95% confidence interval for the mean of the entire senior class.
(b) How large a sample is required in (a) if we want to be 95% confident of is off by
less than 0.05?
Solution
Let be the mean of the entire senior class.
Given: n = 36, 6 . 2 = x , s = 0.3, (1 - o ) = 0.95 05 . 0 = o
(a) A 95% confidence interval estimate for the is

|
|
.
|

\
|
+ < <
|
|
.
|

\
|
+ < <
36
3 . 0
96 . 1 6 . 2
36
3 . 0
96 . 1 6 . 2

025 . 0 025 . 0

o

o
n
z x
n
z x
698 . 2 502 . 2 < <
(b) Let
1
n be the required sample size.
To be 95% confident that if off by less than 0.05 would imply
05 . 0
3 . 0
96 . 1 05 . 0

1 1
025 . 0
s
|
|
.
|

\
|
s
|
|
.
|

\
|
n n
z
o

139 30 . 138
05 . 0
) 3 . 0 )( 96 . 1 (
2
1
~ =
(

> n

86

Note
The following table gives the summary for constructing (1 o)% confidence interval for
mean (single mean and difference of means) for large and small samples
Estimating Conditions Formula
Mean
Large samples (n > 30)
OR o is known
n
Z X
n
Z X
o

o
o o
2 2
+ < <
Mean
Small samples and o
unknown
n
s
t X
n
s
t X
2
,
2
,
o
v
o
v
+ < <
v=n-1
Difference of means
Large sample OR
1
o and
2
o are known
2
2
1
2
1
2
2 1

) (
n n
Z X X
o o
o
+
Difference of means
Small sample &
1
o and
2
o are unknown, assume
2 1
o o =
2 1
2
,
2 1
1 1
) ( ) (
n n
s t X X
p
+
o
v

2
2 1
+ = n n v ,
pooled estimate of sample
standard deviation:
2
) 1 ( ) 1 (
2 1
2
2 2
2
1 1
+
+
=
n n
s n s n
s
p

Difference of means
Small sample &
1
o and
2
o are unknown,
assume
1
o =
2
o

2 2
1 2
1 2
,
2
1 2
( )
s s
X X t
n n
o
v
+
1
) / (
1
) / (
) / / (
2
2
2
2
2
1
2
1
2
1
2
2
2
2 1
2
1

+
=
n
n s
n
n s
n s n s
u
Difference of means Paired observations
, / 2
d
v
s
d t
n
o
; 1 2 d x x =
and v=n-1

Example 4.5
The contents of seven similar containers of sulfuric acid are 9.8, 10.2, 10.4, 9.8, 10.0,
10.2 and 9.6 liters. Find a 95% confidence interval for the mean of all such containers,
assuming an approximate normal distribution.

Solution
Let be the mean of all such containers.
Given: n = 7

= 70 x

= 48 . 700
2
x
87

10
7
70
= = =

n
x
x 08 . 0
6
7
70
48 . 700
1
) (
2
2
2
2
=

n
n
x
x
s
s = 2828 . 0 08 . 0 = ; (1 - o ) = 0.95 05 . 0 = o , 447 . 2
025 . 0 , 6
= t
A 95% confidence interval estimate for the is

|
|
.
|

\
|
+ < <
|
|
.
|

\
|
+ < <
7
2828 . 0
447 . 2 10
7
2828 . 0
447 . 2 10
025 . 0 , 6 025 . 0 , 6

n
s
t x
n
s
t x
262 . 10 738 . 9 < <

Example 4.6
A standardized chemistry test was given to 50 girls and 75 boys. The girls made an
average grade of 76 with a standard deviation of 6, while the boys made an average
grade of 82 with a standard deviation of 8. Find a 96% confidence interval for the
difference
1
and
2
, where
1
is the mean score of all boys and
2
is the mean score of
all girls who might take this test.

Solution
Given: 75
1
= n , 50
2
= n , 82
1
= x , 8
1
= s , 76
2
= x , 6
2
= s ,
04 . 0 96 . ) 1 ( = = o o
A 96% confidence interval for
2 1
is:
2
2
2
1
2
1
02 . 0 2 1 2 1
2
2
2
1
2
1
02 . 0 2 1

) (

) (
n n
z x x
n n
z x x
o o

o o
+ + < < +
50
6
75
8
05 . 2 ) 76 82 (
50
6
75
8
05 . 2 ) 76 82 (
2 2
2 1
2 2
+ + < < + ,
30
1
> n & 30
2
> n , so
1 1
s = o &
2 2
s = o
57 . 8 43 . 3
2 1
< <

Example 4.7
In a batch chemical process, two catalysts are being compared for their effect on the
output of the process reaction. A sample of 12 batches is prepared using catalyst 1 and
a sample of 10 batches was obtained using catalyst 2. The 12 batches for which catalyst
1 was used gave an average yield of 85 with a sample standard deviation of 4, while the
88

average for the second sample gave an average of 81 and a sample standard deviation of
5. Find a 90% confidence interval for the difference between the population means,
assuming the populations are approximately normally distributed with equal variances.

Solution
Let
1
and
2
be the mean population yield using catalyst 1 and catalyst 2, respectively.
Given: 12
1
= n , 10
2
= n , 85
1
= x , 4
1
= s , 81
2
= x , 5
2
= s ,
10 . 0 90 . ) 1 ( = = o o , 20 2 10 12 2
2 1
= + = + = n n v , 725 . 1
05 . 0 , 20
= t
pooled estimate of sample standard deviation
2
) 1 ( ) 1 (
2 1
2
2 2
2
1 1
+
+
=
n n
s n s n
s
p

478 . 4
2 10 12
5 ) 1 10 ( 4 ) 1 12 (
2 2
=
+
+
=
A 90% confidence interval for
2 1
is:
2 1
05 . 0 , 20 2 1 2 1
2 1
05 . 0 , 20 2 1
1 1
) ( ) (
1 1
) ( ) (
n n
s t x x
n n
s t x x
p p
+ + < < +
10
1
12
1
) 478 . 4 )( 725 . 1 ( ) 81 85 (
10
1
12
1
) 478 . 4 )( 725 . 1 ( ) 81 85 (
2 1
+ + < < +
31 . 7 69 . 0
2 1
< <

Example 4.8
The weight of 10 adults selected randomly before and after a certain new diet was
introduced was recorded as follows:
1
x ) After (
2
x ) Difference
1 76 81 -5
2 60 52 8
3 85 87 -2
4 58 70 -12
5 91 86 5
6 75 77 -2
7 82 90 -8
8 64 63 1
9 79 85 -6
10 88 83 5
Find a 98% confidence interval for the mean difference in weight.
89

Solution
i
d
d
n
=

= -1.6
2
2
( ( 1.6))
40.7
1
i
d
d
s
n

= =

For v = n-1 = 9;
0.01
2.821 t = .
A 98% confidence interval is
6.38
1.6 (2,821)
10
| |

|
\ .

That is 7.29 4.09
d
< <

Exercise 4.1
4.1.1 What is the value of the finite population correction factor for

when (i) n = 10
and N = 1000; (ii) n = 5 and N = 200?
4.1.2 How many different samples of size n=2 can be chosen from a finite population of
size (i) N=25; (ii) N=6?
4.1.3 When we sample from an infinite population, what happens to the standard error
of the mean if the sample size is (a) increased from 50 to 200 (b) decreased from
225 to 25?
4.1.4 The mean of a random sample size n=25 is used to estimate the mean of an infinite
population with the s.d. o = 2.4. What can we assert about the probability that the
error will be less than 1.2, if we use the central limit theorem?
4.1.5 A random sample of size 100 is taken from an infinite population having the mean
= 76 and the variance of o
2
= 256. What is the probability that x
will be between
75 and 78?

4.1.6 A random sample of size n = 36 is taken from an infinite population with mean =
63 and the variance of o
2
= 81. What can we assert about the probability of getting
a sample mean greater than 66.75, if we use the central limit theorem?
4.1.7 A research worker wants to determine the average time it takes a mechanic to
rotate the tires of a car and she wants to be able to assert with 95% confidence
that the mean of her sample is off by at most 0.50 minute. If she can presume
from past experience that o = 1.6 minutes, how large a sample will she have to
take?
90

4.1.8 If we want to determine the average mechanical aptitude of a large group of
workers, how large a random sample will we need to be able to assert with
probability 0.95 that the sample mean will not differ from the true mean by more
than 3.0 points? Assume that it is known from past experience that o = 20.0.
4.1.9 What is the maximum error one can expect to make with probability 0.90 when
using the mean of a random sample of size n = 64 to estimate the mean of a
population with o
2
= 2.56?
4.1.10 A random sample of 40 drums of a chemical, drawn from among 200 such drums
whose weights can be expected to have the o = 12.2, has a mean weight of 240.8
pounds. If we estimate the mean weight of all 200 drums as 240.8 pounds, what
can we assert with 99% confidence about the maximum error?
4.1.11 To estimate the average time it takes to assemble a certain computer
component, the industrial engineer at an electronics firm timed 40 technicians in
the performance of this task, getting a mean of 12.73 minutes with a standard
deviation of 2.06 minutes. Construct a 98% confidence interval for the true
average time it takes to assemble the computer component.
4.1.12 Ten bearings made by certain process have a mean diameter of 0.5060 cm with a
standard deviation of 0.0040 cm. Assuming that the data may be looked upon as a
random sample from a normal population, construct 95% and 99% confidence
interval for the actual average diameter of bearings made by this process.

4.10 Statistical Hypothesis
Any statement regarding the characteristics of the population or the parameter of the
population or the nature of the population probability distribution is known as a
Hypothesis.

For example: i) the ratio of women for every 1000 men is 937 in India, ii) the literacy rate
in India is 56.7%, iii) there is substantial growth in prices of the infrastructure goods in
the past five years in India., and so on.
Further the hypothesis is classified into two types namely: Simple hypothesis any
hypothesis which specifies the population completely or absolutely is known as simple
hypothesis, otherwise the hypothesis is said to be Composite or Comparative
91

hypothesis. In the above example (i) & (ii) are the simple hypothesis and (iii) is the
composite hypothesis.

4.11 Test of Hypothesis
The Test of Hypothesis is nothing but a two-action decision problem, either to accept
the hypothesis or to reject the hypothesis.

4.12 Test of Significance
The test of significance is to test the deviation between
i) the parameter and statistic, or
ii) the difference between two independent statistics, is just because of
chance variation or assignable variation.

4.13 Null Hypothesis
Any statement or hypothesis, which gives no difference between the parameter and a
statistic or no differences in two independent statistics, is known as Null Hypothesis.

According to Prof. Fisher, Null hypothesis is a hypothesis, which is tested for possible
rejection under the assumption that it is true. The null hypothesis is always denoted by H
0
:

4.14 Alternative Hypothesis
Any statement or hypothesis, which gives a difference between the parameter and a
statistic or difference in between two independent statistics, is known as an alternative
hypothesis. The alternative hypothesis is always denoted by H
1

4.15 Errors in Sampling
There are four possible situations in any testing of hypothesis procedure and are given in
the following table:

0
H is correct
0
H is wrong
Accept
0
H Correct decision Type II Error
Reject
0
H Type II error Correct decision

92

In the above table, there are two types of errors occurred in sampling, while testing an
hypothesis problem, which are given by
- Type I Error: Rejecting the Null Hypothesis H
0
, when it is true.
- Type II Error: Accepting the Null Hypothesis H
0
, when it is false.

4.16 Sizes of Errors
The probabilities of occurring the errors are known as size of the errors and are
respectively given by:
Size of Type-I Error = Prob. {Type I Error} = Prob. {Reject H
0
| H
0
} = o
Size of Type-II Error = Prob. {Type II Error} = Prob. {Accept H
0
| H
1
} = |
Here o, is known as Level of significance and |, is known as power of a test.

4.17 Critical Region and Critical Value
The region of rejecting the null hypothesis is known as critical region and is denoted by
o. Generally, the critical region is also known as the level of significance, and is fixed in
advance and is either 5% or 1%.
A value, which separates out the acceptance region and rejection region, is known as
critical value.

4.17.1 Significant or Critical Values
The following table represents the significant values at different levels of significance.
Z Test
Level of Significance (o%)
1% 5% 10%
Two-tailed Test 2.58 1.96 1.645
Right-tailed Test 2.33 1.645 1.25
Left-tailed Test -2.33 -1.645 -1.25

Note:
The significant values for small sampling techniques ( i.e. using t-Test, F-Test and Chi-
Square (_
2
) Test) depends upon the size of the sample.

93

4.18 Hypothesis Testing Terminology
1. Null hypothesis, H
0
: A hypothesis that is held to be true until very strong
evidence to the contrary is obtained. H
0
:
0
=
2. Alternative hypothesis, H
1
: It is a hypothesis that is complement to the null
hypothesis. Hence it will be accepted if the null hypothesis is rejected.
1 0
: H = (two-tail test)
1 0
: H > (One-tail test)
1 0
: H < (One-tail test)
In the one-tail test we have some expectation about the direction of the error
when the null hypothesis is wrong, while in the two-tail test we dont have such
expectation.
3. Test statistics is the value, based on the sample, used to determine whether the
null hypothesis should be rejected or accepted.
4. Critical region is a region in which if the test statistic falls the null hypothesis will
be rejected.
5. Types of error
(a) Type I error: Reject H
0
when H
0
is true
(b) Type II error: Accept H
0
when H
a
is true
6. The significance level, o is the probability of committing a Type I error, i.e.,
P(Type I error) = o.
The probability of committing a Type II error is |; i.e., P(Type II error) = |.

4.19 Basic Steps in Testing Hypothesis
1. Formulate the null hypothesis.
2. Formulate the alternative hypothesis.
3. Specify the level of significance to be used.
4. Select the appropriate test statistic and establish the critical region.
5. Compute the value of the test statistic.
6. Conclusion: Reject H
0
if the statistic has a value in the critical region, otherwise
accept H
0
.

94

4.20 Tests Concerning Means
The tests concerning means (for large and small samples) are summarized in the
following table.
H
0
Conditions Test statistic

0
=
Large samples (n > 30) OR o
is known
n
x
z
o

0

=
Small samples and o
unknown
n s
x
t
0

= with 1 = n u

0 2 1
d = Large samples OR
1
o and
2
o are known
2
2
2
1
2
1
0 2 1
) (
n n
d x x
z
o o
+

=

Small sample &
1
o and
2
o
are unknown, assume
2 1
o o =
|
|
.
|

\
|
+

=
2 1
0 2 1
1 1
) (
n n
s
d x x
t
p

with 2
2 1
+ = n n u and
2
) 1 ( ) 1 (
2 1
2
2 2
2
1 1 2
+
+
=
n n
s n s n
s
p

if
1
o =
2
o but unknown

Small sample &
1
o and
2
o
are unknown,
assume
1
o =
2
o

2
2
2
1
2
1
0 2 1
) (
n
s
n
s
d x x
t
+

=
with
1
) / (
1
) / (
) / / (
2
2
2
2
2
1
2
1
2
1
2
2
2
2 1
2
1

+
=
n
n s
n
n s
n s n s
u

0 2 1
d =
Paired observations
n s
d d
t
d
0

= with 1 = n u

Example 4.9
A manufacturer of sports equipment has developed a new synthetic fishing line that he
claims has a mean breaking strength of 8 kilograms with a standard deviation of 0.5
kilogram. Test the hypothesis that = 8 kilograms against the alternative that = 8
kilograms if a random sample of 50 lines is tested and found to have a mean breaking
strength of 7.8 kilograms. Use a 0.01 level of significance.

95

Solution
Null hypothesis: 8 = kilograms
Alternative hypothesis: 8 = kilograms (Two tailed test alternative)
Level of significance: 0.01
Critical region: Z > 58 . 2
005 . 0
= z or Z < 58 . 2
005 . 0
= z
Computation:

n = 50 8 . 7 = x 5 . 0 = o

828 . 2
50
5 . 0
8 8 . 7
=

=
n
x
z
o

Conclusion: As the sample z (= -2.828) falls inside the critical region, so reject the null
hypothesis at 0.01 level of significance and conclude that is significantly smaller than
8 kilograms.

Example 4.10
The average length of time for students to register for fall classes at a certain college has
been 50 minutes with a standard deviation of 10 minutes. A new registration procedure
using modern computing machines is being tried. If a random sample of 12 students had
an average registration time of 42 minutes with a standard deviation of 11.9 minutes
under the new system, test the hypothesis that the population mean is now less than
50, using a level of significance of (1) 0.05, and (2) 0.01. Assume the population of times
to be normal.

Solution
Let be the population mean time for students to register in the new registration
procedure.
Null hypothesis: 50 = minutes
Alternative hypothesis: 50 < minutes (left tailed test alternative)
Level of significance: 0.05
Critical region: (n = 12 < 30; and the new o is unknown, so t-test should be used)
degree of freedom (v ) = n -1 = 12 -1 =11
96

t < 796 . 1
05 . 0 , 11
= t
Computation:
n = 12 42 = x s = 11.9
329 . 2
12
9 . 11
50 42
=

=
n
s
x
t

Conclusion:
(1) As sample t (= -2.329) falls inside the critical region, so reject the null hypothesis
at 0.05 level of significance and conclude that is significantly smaller than 50
minutes.
(2) Identical with those of (1) except the critical region would be replaced by:
718 . 2
01 . 0 , 11
= < t t
and the corresponding conclusion would be changed as follows:
As sample t (= -2.329) falls outside the critical region, so reject the alternative
hypothesis at 0.01 level of significance and conclude that is not highly
significantly smaller than 50 minutes.

Example 4.11
An experiment was performed to compare the abrasive wear of two different laminated
materials. Twelve pieces of material 1 were tested, by exposing each piece to a machine
measuring wear. Ten pieces of material 2 were similarly tested. In each case, the depth
of wear was observed. The samples of material 1 gave an average (coded) wear of 85
units with a standard deviation of 4, while the samples of material 2 gave an average of
81 and a standard deviation of 5. Test the hypothesis that the two types of material
exhibit the same mean abrasive wear at the 0.10 level of significance. Assume the
populations to be approximately normal with equal variances.

Solution:
Let
1
and
2
be the mean abrasive wear of material 1 and 2 respectively.
Null hypothesis:
2 1
= , i.e. 0
2 1
=
Alternative hypothesis:
2 1
= , i.e. 0
2 1
=
Level of significance: 0.10
97

Critical region: (As both
1
n and
2
n are smaller than 30 and their standard deviations are
unknown, so t-test has to be used.)
20 2 10 12 2
2 1
= + = + = n n v , 725 . 1
05 . 0 , 20
= > t t or 725 . 1
05 . 0 , 20
= < t t
Computation:
12
1
= n 85
1
= x 4
1
= s
10
2
= n 81
2
= x 5
2
= s
05 . 20
2 10 12
5 ) 9 ( 4 ) 11 (
2
) 1 ( ) 1 (
2 2
2 1
2
2 2
2
1 1 2
=
+
+
=
+
+
=
n n
s n s n
s
p

086 . 2
10
1
12
1
05 . 20
0 ) 81 85 (
1 1
) ( ) (
2 1
2 1 2 1
=
+

=
+

=
n n
s
x x
t
p

Conclusion: As the sample t (=2.086) falls inside the critical region, so reject the null
hypothesis at 0.10 level of significance and conclude that the mean abrasive wear of
material 1 is significantly higher than that of the material 2.

Example 4.12
Five samples of a ferrous-type substance are to be used to determine if there is a
difference between a laboratory chemical analysis and an X-ray fluorescence analysis of
the iron content. Each sample was split into two sub-samples and the two types of
analysis were applied. Following are the coded data showing the iron content analysis:
Sample
Analysis 1 2 3 4 5
x-ray 2.0 2.0 2.3 2.1 2.4
Chemical 2.2 1.9 2.5 2.3 2.4
Solution:
Assuming the populations normal, test at the 0.05 level of significance whether the two
methods of analysis give, on the average, the same result.

Let
1
and
2
be the mean iron content determined by the laboratory chemical
analysis and X-ray fluorescence analysis respectively; and

D
be the mean of the population of differences of paired measurements.

98

Null hypothesis:
2 1
= or 0 =
D

Alternative hypothesis:
2 1
= or 0 =
D

Level of significance: 0.05
Critical region: (As n = 5 < 30, so t-test should be used.)
776 . 2
025 . 0 , 4
= > t t or 776 . 2
025 . 0 , 4
= < t t
Computation:
Sample
Analysis 1 2 3 4 5
x-ray 2.0 2.0 2.3 2.1 2.4
Chemical 2.2 1.9 2.5 2.3 2.4
i
d
-0.2 0.1 -0.2 -0.2 0
2
i
d
0.04 0.01 0.04 0.04 0
5 . 0
5
1
=

= i
i
d 13 . 0
5
1
2
=

= i
i
d
1 . 0
5
5 . 0
5
=

= =

d
d

( )
02 . 0
) 4 )( 5 (
) 5 . 0 ( ) 13 . 0 )( 5 (
) 1 (
2
2
2
2
=

=

=

n n
d d n
s
d

5811 . 1
5
02 . 0
0 ) 1 . 0 (
=

=

=
n
s
d
t
d
D

Conclusion: As the sample t (=-1.5811) falls outside the critical region, so reject the
alternative hypothesis at 0.05 level of significance and conclude that there is no
significant difference in the mean iron content determined by the above two analyses.

4.21 Chi-square Tests
The various tests of significance studied earlier such that as Z-test, t-test, F-test was
based on the assumption that the samples were drawn from normal population. Under
this assumption the various statistics were normally distributed. Since the procedure of
testing the significance requires the knowledge about the type of population or
parameters of population from which random samples have been drawn, these tests are
99

known as parametric tests.

But there are many practical situations in which the assumption of any kind about the
distribution of population or its parameter is not possible to make. The alternative
population is made is known as non-parametric tests. Chi-square test is an example of
the non parametric test. Chi-square distribution is a distribution free test.

Chi-square distribution was first discovered by Helmert in 1876 and later independently
by Karl Pearson in 1900. The range of chi-square distribution is 0 to .

The square of a standard normal variate is known as chi-square variate with one degree
of freedom. Symbolically we can write it as

test is used for enumeration data (the data obtained by enumeration or counting is
called enumeration data. For example, number of blue flowers, number of intelligent
boys, number of curled leaves, etc.,) which generally relate to discrete variable where as
t-test and standard normal deviate tests are used for measure mental data which
generally relate to continuous variable.

test can be used to know whether the given objects are segregating in a theoretical
ratio or whether the two attributes are independent in a contingency table.

a) Conditions for the validity of

test:
The validity of

-test of goodness of fit between theoretical and observed, the

following conditions must be satisfied.
i) The sample observations should be independent
ii) Constraints on the cell frequencies, if any, should be linear o
i
=e
i

iii) N, the total frequency should be reasonably large, say greater than 50
100

iv) If any theoretical (expected) cell frequency is < 5, then for the application of chi-
square test it is pooled with the preceding or succeeding frequency so that the pooled
frequency is more than 5 and finally adjust for the d.f. lost in pooling.

b) Applications of Chi-square Test:
1. testing the independence of attributes
2. to test the goodness of fit
3. testing of linkage in genetic problems
4. comparison of sample variance with population variance
5. testing the homogeneity of variances
6. testing the homogeneity of correlation coefficient

4.21.1 Goodness-of-fit Test
A test to determine if a population has a specified theoretical distribution. The test is
based on how good a fit we have between the frequency of occurrence of observations
in an observed sample and the expected frequencies obtained from the hypothesized
distribution.

A goodness-of-fit test between observed and expected frequencies is based on the
quantity _
test
2
2
=

( ) O E
E
i i
i

where _
test
2
is a value of the random variable whose sampling distribution is
approximated very closed by the Chi-square distribution,
O
i
is the observed frequency of cell i,
E
i
is the expected frequency of cell i.

The number of degrees of freedom in a Chi-square goodness-of-fit test is equal to the
number of cells minus the number of quantities obtained from the observed data that
are used in the calculations of the expected frequencies.
For a level of significance equal to o _ _
o
,
test
2 2
> constitutes the critical region. The
decision criterion described here should not be used unless each of the expected
frequencies is at least equal to 5.
101

Example 4.13
Consider the tossing of a die 120 times.

Faces
1 2 3 4 5 6
Observed 20 22 17 18 19 24
Expected

By comparing the observed frequencies with the expected frequencies, one has to
decide whether the die is fair die or not.
Null hypothesis: the die is a fair die, i.e.
6
1
) ( = = i X P for i = 1, 2, 3, 4, 5, and 6
Alternative hypothesis: the die is not a fair die
Level of significance: 0.05
Critical region: 5 1 6 1 = = = n v ; 07 . 11
2
05 . 0 , 5
2
= > _ _
Computation:
Expected value = 20 )
6
1
( 120 ) ( = = = i X nP

I 1 2 3 4 5 6
Observed ) (
i
O
20 22 17 18 19 24
Expected ) (
i
E
20 20 20 20 20 20
i i
E O
0 2 -3 -2 -1 4

=
6
1
2
2
) (
i i
i i
E
E O
_

7 . 1
20
4
20
) 1 (
20
) 2 (
20
) 3 (
20
2
20
0
2 2 2 2 2 2
= +

+ + =

Conclusion: As the sample
2
_ (= 1.7) falls outside the critical region, so reject the
alternative hypothesis and conclude that the die is a fair die.

4.21.2 Test for Independence
The Chi-square test procedure can also be used to test the hypothesis of independence
of two variables/attributes. The observed frequencies of two variables are entered in a
two-way classification table, or contingency table.
The expected frequency of the cell in the i
th
row and j
th
column in the contingency table.
102

The degrees of freedom for the contingency table is equal to (r 1) (c 1) where r is the
number of rows and c is the number of columns in the table.

Example 4.14
Suppose that we wish to study the relationship between grade point average and
appearance of 150 students of a college, whose analysis are given below. Use the level
of significance is 0.05
Appearance 1 2 3 4 Totals
Attractive 14 ( ) 11 ( ) 10 ( ) 5 ( ) 40
Ordinary 10 ( ) 16 ( ) 16 ( ) 14 ( ) 56
Unattractive 3 ( ) 4 ( ) 7 ( ) 10 ( ) 24
Totals 27 31 33 29 120

Null hypothesis: There is no relationship between grade point average and appearance.
That is, the two characteristics are independent.
Alternative hypothesis: There is a relationship between grade point average and
appearance. That is, the two characteristics are not independent.
Level of significance: 0.05
Critical region:
2
05 . 0 ,
2
v
_ _ > , where v = (r -1)(c - 1)
Computation:
Appearance 1 2 3 4 Totals
Attractive 14
(9)
11
(10.33)
10
(11)
5
(9.67)
40
Ordinary 10
( 12.6)
16
( 14.47)
16
(15.4)
14
(13.53)
56
Unattractive 3
(5.4)
4
(6.2)
7
(6.6)
10
(5.8)
24
Totals 27 31 33 29 120

67 . 9
) 67 . 9 5 (
11
) 11 10 (
33 . 10
) 33 . 10 11 (
9
) 9 14 ( ) (
2 2 2 2 2
2

+

E
E O
_

53 . 13
) 53 . 13 14 (
4 . 15
) 4 . 15 16 (
47 . 14
) 47 . 14 16 (
6 . 12
) 6 . 12 10 (
2 2 2 2

+
103

818 . 10
8 . 5
) 8 . 5 10 (
6 . 6
) 6 . 6 7 (
2 . 6
) 2 . 6 4 (
4 . 5
) 4 . 5 3 (
2 2 2 2
=

+

Conclusion: Since 596 . 12
2
05 . 0 , 6
= _ , so sample
2
_ (=10.818) falls outside the critical
region. So reject the alternative hypothesis and conclude that there is no evidence to
support there is relationship between grade point average and appearance.

4.21.3 Test for Homogeneity
To test the hypothesis that several population proportions are equal.
The approach for the test of homogeneity is the same as for the test of independence of
variables/attributes.

Example 4.15
A study of the purchase decisions for 3 stock portfolio managers, A, B, and C was
conducted to compare the rates of stock purchases that resulted in profits over a time
period that was less than or equal to one year. One hundred randomly selected
purchases obtained for each of the managers showed the results given in the table. Do
these data provide evidence of differences among the rates of successful purchases for
the three portfolio managers? Test with . 05 . 0 = o

Result
Manager
A B C
Purchases show profit 63 71 55
Purchase do no show profit 37 29 45
Total 100 100 100

Null hypothesis: the rates of stock purchases that resulted in profit were the same for
the three stock portfolio managers
Alternative: their rates were not all the same
Level of significance: 0.05
Critical region: 2 ) 1 3 )( 1 2 ( = = v ; 991 . 5
2
05 . 0 , 2
2
= > _ _
Computation:

104

Result
Manager
A B C Total
Purchases show profit
63
(63)
71
(63)
55
(63)
189
Purchase do no show profit
37
(37)
29
(37)
45
(37)
111
Total 100 100 100 300

49 . 5
37
) 37 45 (
37
) 37 29 (
37
) 37 37 (
63
) 63 55 (
63
) 63 71 (
63
) 63 63 (
2 2 2 2 2 2
2
=

= _

Conclusion: As the sample
2
_ (= 5.49) falls outside the critical region so reject the
alternative hypothesis and conclude that there is no sufficient evidence to support the
rates of purchases resulted in profit of the three portfolio managers were different.

Exercise 4.2
4.2.1 According to the norms established for a mechanical aptitude test, persons who
are 18 years old should average 73.2 with standard deviation of 8.6. If 45
randomly selected persons of that age averaged 76.7, test the null hypothesis =
73.2 against the alternative hypothesis > 73.2 at the 0.01 level of significance.
4.2.2 In 64 randomly selected hours of production, the mean and standard deviation of
the number of acceptable pieces produced by an automatic stamping machine is
1,038 and 146 respectively. At the 0.05 level of significance does this enable us to
reject the null hypothesis = 1000 against the alternative hypothesis >1000?
4.2.3 A random sample of 6 steel beams has a mean compressive strength of 58,392psi,
with a standard deviation of 648psi. Use this information and the level of
significance of o = 0.05 to test whether the true average compressive strength of
the steel beam from which this sample came is 58,000psi. Assume normality.
4.2.4 Five measurements of tar content of a certain kind of cigarette yielded 14.5, 14.2,
14.4, 14.3 and 14.6 mg per cigarette. Show that the difference between the mean
of this sample, = 14.4 and the average tar claimed by the manufacturer = 14.0,
is significant at o = 0.05. Assume normality.
4.2.5 A manufacturer of gunpowder has developed a new powder which is designed to
produce a muzzle velocity equal to 3000 ft./sec. Eight shells are loaded with the
charge and the muzzle velocities measured. The resulting velocities are as: 3005,
105

2935, 2965, 2995, 3905, 2935 and 2905. Do these data present sufficient evidence
to indicate that the average velocity differs from 3000ft/sec?
4.2.6 An investigation of two kinds of photocopying equipment showed that 71 failures
of the first kind of equipment took on the average 83.2 minutes to repair with a
standard deviation of 19.3 minutes, while 75 failures of the second kind of
equipment took on the average 90.8 minutes to repair with a standard deviation of
21.4 minutes. Test the null hypothesis

(the hypothesis that on the

average it takes an equal amount of time o repair either kind of equipment)
against the alternative hypothesis

at the level of significance of o =

0.05
4.2.7 Suppose that we want to investigate whether on the average men earn more than
\$20 per week more than women in a certain industry. If sample data show that 60
men earn on the average

per week with a standard deviation of s

1
=
\$15.60, while 60 women earn on the average

per week with a

standard deviation of s
2
= \$18.20, what can we conclude at the 0.01 level of
significance?
4.2.8 Measuring specimens of nylon yarn taken from two spinning machines, it was
found that 8 specimens from the first machine had a mean denier of 9.67 with a
standard deviation of 1.81, while 10 specimens from the second machine has a
mean denier of 7.43 with a standard deviation of 1.48. Assuming that the
populations sampled are normal and have the same variance, test the null
hypothesis

against the alternative hypothesis

at the
level of significance of o = 0.05.
4.2.9 The following are the Brinell hardness values obtained for samples of two
magnesium alloys:
Alloy 1: 66.3 63.5 64.9 61.8 64.3 64.7 65.1 64.5 68.4 63.2
Alloy 2: 71.3 60.4 62.6 63.9 68.8 70.1 64.8 68.9 65.8 66.2
Use the 0.05 level of significance to test null hypothesis

against the
alternative hypothesis

.
4.2.10 A genetical law says that children having one parent of blood group M and other
parent of blood group N will always be one of the three groups M, MN and N;
and that the average number of children in these three groups will be in the ratio
106

1:2:1. The report on an experiment states as follows: Of 162 children having
one M Parent, and one N parent, 28% were found to be of group M, 42% of
group MN and the rest of them are group N. Do the data in the report conform
to the expected genetic ratio 1:2:1? (Given value of Chi-square at 2, 0.05 is 5.991)
4.2.11 A chemical extraction plant processes sea water to collect sodium chloride and
magnesium. It is known that sea water contains sodium chloride, magnesium
and other elements in the ratio of 62:4:34. A sample of 200 tones of sea water
has resulted in 130 tones of sodium chloride, 6 tones of magnesium and 64 tones
of other elements. Are these data consistent with the known composition of sea
water at 5% level?
4.2.12 The following table shows the number of aircraft accidents that occurred during
the six days of a week. Find whether the accidents are uniformly distributed over
the week? Given value of _
2
= 11.070 with 2, 0.05
Day Mon Tue Wed Thu Fri Sat Total
No. of
Accidents
14 18 12 11 15 14 84

4.2.13 200 digits were chosen at random from a set of tables. The frequencies of the
digits were as follows:
Digit 0 1 2 3 4 5 6 7 8 9 Total
F 18 19 23 21 16 25 22 20 21 15 200
Use Chi-square test; to assess the correctness of the hypothesis that the digits
were distributed in equal number in the table, given the value of chi-square at 9,
0.05 is 16.9
4.2.14 In an experiment on immunization of cattle from tuberculosis, the following
results were obtained:
Affected Unaffected
Inoculated 12 28
Not Inoculated 13 7
Examine the effect of vaccine in controlling the incidence of the disease.
4.2.15 Can vaccination be regarded as a preventive measure of small-pox as evidenced
by the following data: Of 1,482 persons exposed to small pox in a locality, 368 in
107

all were attacked. Of these 1,482 persons, 343 were vaccinated and of these,
only 35 were attacked.
4.2.16 Out of 8000 graduates in a city 800 are females; out of 1600 graduate employees
120 are females. Use _
2
test to determine if any distinction is made in
appointment on the basis of gender. (Value of _
2
is 3.84 for 1d.f., at 0.05 level)