STATISTICAL HYPOTHESIS
A statistical hypothesis is an assertion regarding the statistical distribution of the
population. It is a statement regarding the parameters of the population
Statistical hypothesis is denoted by H
Examples:
1. H: The population has mean μ = 25
2. H: The population is normally distributed with mean μ=25 and standard deviation
In a test procedure, to start with, a hypothesis is made. The validity of this hypothesis is tested. If
the hypothesis is found to be true, it is accepted. On the other hand, if it is found to be untrue, it
is rejected
The hypothesis which is being tested for possible rejection is called null hypothesis. The null
hypothesis is denoted by H0. Hypothesis which is accepted when the null hypothesis is rejected is
called alternative hypothesis The alternative hypothesis is denoted by H1.
CRITICAL REGION
From a population many samples of the same size n can be drawn. Let S be the set of all such
sample of size n that can be drawn from the population. Then, S is called sample space. While
testing a null hypothesis, among the samples which belong to S, some samples lead to the
acceptance of the null hypothesis, whereas, some others lead to the rejection of the null hypothesis.
The set of all those samples belonging to the sample space which lead to the rejection of the
null hypothesis is called critical region. The critical region ids denoted by ω. The critical region
is also rejection region. The set of samples which lead to the acceptance of the null hypothesis is
the acceptance region. It is (S- ω).
In fact, to decide whether the sample in hand belongs to ω, the criterian |Z| > k is adopted. And
so, in effect, the critical region is defined by |Z| >ks.
ERRORS OF THE FIRST AND THE SECOND KIND
Type I and type II errors)
While testing a null hypothesis against an alternative hypothesis, one f the following four
situations arise
Actual fact Decision based on the Error
sample
1 H0 is true accept H0 correct decision -
2 H0 is true reject H0 wrong decision Type I
3 H0 is not true accept H0 wrong decision Type II
4 H0 is not true reject H0 correct decision -
Here, in situations (2) and (3), wrong decisions are arrived at. These wrong decisions are Error of
the first kind (Type I error) and Error of the second kind (Type II error) respectively. Thus,
(i) Error of the first kind (Type I error) is taking a wrong decision to reject the null
hypothesis when it is actually true.
(ii) Error of the second kind (Type II error) is taking a wrong decision to accept the null
hypothesis when it is actually not true.
The probability of occurrence of the first kind of error is denoted by α. It is called level of
significance. Thus, the level of significance is the probability of Type I error. It is the
probability of rejection of the null hypothesis when it is actually true. Usually the level of
significance is fixed at 0.05 or 0.01. In other words, the level is fixed at 5% or 1%.
The probability of occurrence of the second kind of error is denoted by β.
The value (1- β) is called power of the test. Power of a test is the probability of rejecting H0
when it is not true.
While testing, the level of significance α is decided in advance. Then, the critical value k is
determined in such a way that the power (1- β) is maximum.
Thus, the critical value k is based on the level of significance. For tests which are based on normal
distribution, if α=0.05, the critical value is k = 1.96. If α=0.01, the critical value is k = 2.58.
Note: In fact, a decision to accepts H0, is based only on the given data. And so, rather than
making an assertive statement ‘H0 is accepted’, we would make a statement ‘H0 is not
rejected’. However, at the level of this book, we will not bother about the subtle difference
between these statements
TWO-TAILED AND ONE-TAILED TESTS (Two-sided and one-sided testes)
While testing H0, if the critical region is considered at one tail of the sampling distribution
of the test statistic, the test is one-tailed test
One the other hand, if the critical region is considered at both
the tails of the sampling distribution of the test statistic,
the test is two-tailed test.
Session 09
TEST FOR MEAN
Suppose the mean of a population is known. We want to test whether the mean is a given value
µ0. The null hypothesis is H0: µ=µ0. The alternative hypothesis is H1:µ≠µ0.
For a large random sample of size n from the population, under H0, the distribution of
x − µ0
Z= is N(0,1)
σ
n
The test statistic is x − µ0
Z =
σ/ n
For the sample, if the calculated value |Z| cal>k, H0 is rejected. On the other hand, if |Z|cal≤ k, H0
is accepted.
For the level of significance α=0.05, the critical value is k = 1.96. However, for α=0.01, the critical
value is k = 2.58.
Note: Here, if σ is not known, the test statistic is Z = | x − µ 0 | where s is the sample standard
σ
n
deviation
TEST FOR EQUALITY OF MEANS
The null hypothesis is H0 : µ1=µ2 (the means of the two populations are equal). The alternative
hypothesis is H1 : µ1≠µ2. Under H0, let µ be the common means and let σ1 and σ2 be the standard
deviations of the two populations.
Let a random sample of size n1 be drawn from the first population. Let the sample be x 1 . Also,
let a random sample of size n2 be drawn from the second population. Let the mean of this sample
be x 2
𝑥𝑥1 −𝑥𝑥2
Then, 𝑍𝑍 = is N(0,1)
𝜎𝜎2 𝜎𝜎2
� 1+ 2
𝑛𝑛1 𝑛𝑛2
|𝑥𝑥1 −𝑥𝑥2 |
And so the test statistic is |𝑍𝑍| =
𝜎𝜎 2
𝜎𝜎 2
� 1+ 2
𝑛𝑛1 𝑛𝑛2
For the samples, if |Z| cal > k, H0 is rejected.
On the other hand, if |Z| cal ≤ k, H0 is accepted. For the level of significance α=0.05, the critical
value is k = 1.96. However, for α=0.01, the critical value is k = 2.58. Here, if σ1 and σ2 are not
|𝑥𝑥1 −𝑥𝑥2 |
known, the test statistic is |𝑍𝑍| = . Where s1 and s2 are the sample standard deviations
𝑠𝑠2 𝑠𝑠2
� 1+ 2
𝑛𝑛1 𝑛𝑛2
TEST FOR PROPORTION
Suppose the proportion of an attribution in a population is not known, we want to test whether
the proportion is a given value P0 . The null hypothesis is H0: P = P0. The alternative hypothesis
is H1: P ≠ P0.
In a large random sample of size n from the population, let x units posses the attribute. Then, the
sample proportion is p=x/n
p − P0
Ans so, Z = is N(0,1).
p 0 Q0
n
p − P0
Therefore, the test statistic is Z =
p 0 Q0
n
For the samples, if |Z|cal > k, H0 is rejected.
On the other hand, if |Z|cal ≤ k, H0 is accepted.
For the level of significance α=0.05, the critical value is k = 1.96. However, for α=0.01, the
critical value is k = 2.58.
TEST FOR EQUALITY PROPORTION
Suppose there are two populations with unknown proportions, and we wish to test whether the
proportions( of certain attributes) in the two populations are equal. The null hypothesis is H0 :
P1=P2 (the proportions are equal). The alternative hypothesis is H1 : P1≠ P2
Under H0, let P be the common proportion. Let a large random sample of size n1 be drawn from
the first population. Among these n1 units, let x1 units possess the attribute, so that the sample
proportion is p1=x1/n. Also let a large random sample of size n2 be drawn from the second
population. Among the units, let x2 units posses the attribute, so that sample proportion is p2 =
x2/n2.
The test statistic is –
Generally, the common proportion P will not be known. And so, it is estimated from the samples.
This estimates is
And also,
For the samples, if |Z|cal > k, H0 is rejected.
On the other hand, if |Z|cal ≤ k, H0 is accepted.
PROBLEMS
1. A random sample of 200 tins of vanaspathi has mean weight 4.97 kgs and standard deviation
0.2kgs. Test at 1% level of significance, that the tins have 5 kgs. Vanaspathi
Solution:
H0 : µ=5kg
H1: µ≠5kg
x − µ0
Z =
σ/ n
4.97 − 5
Under H0 Z =
02 / 200
=2.12
Ztab=2.58 at 1% l.o.s
Since Zcal < Ztab, we accept H0 at 1% l.o.s i.e the tins have 5 kgs of vanaspathi at 1% l.o.s
2. A random sample of 100 rods drawn from a lot of rods has mean length 32.7cms. And
standard deviation 1.3cms. Can it be concluded that the lot has mean 32 cms?
Solution:
H0 : µ=32
H1: µ≠32
x − µ0
Z =
σ/ n
32.7 − 32
Under H0 Z = = 5.38
1.3 / 100
Ztab=1.96 at 5% l.o.s
Since Zcal > Ztab, we reject H0 at 5% l.o.s i.e the lot does not have mean 32 cms
3. The mean and standard deviation of heights of 100 randomly selected boys are 163 cms and 3
cms, respectively. The mean and standard deviation of heights of 80 randomly selected girls are
161 cms and 2 cms, respectively. Can it be concluded at 1% level of significance that boys and
girls are equally tall?
Solution:
H0 : µ1= µ2
H1: µ1≠ µ2
|𝑥𝑥1 − 𝑥𝑥2 |
|𝑍𝑍| =
𝜎𝜎12 𝜎𝜎22
�
𝑛𝑛1 + 𝑛𝑛2
163 − 161
Under H0 Z = = 5.35
32 2 2
+
100 80
Ztab=2.58
Since Zcal > Ztab, we reject H0 at 1% l.o.s i.e boys & girls do not have equal height at 1% l.o.s.
4. The standard deviation of length of fibre manufactured by process A is 0.5cms, and standard
deviation of length of fibre manufactured by process B is 0.6cms. A sample of 40 randomly
selected fibres from process A has mean length 16.7 cms. A sample of 60 randomly selected
fibres from process B has mean length 16.4 cms. Test whether process A and process B differ
with regard to length of fibre manufactured by them.
Solution:
H0 : µ1= µ2
H1: µ1≠ µ2
|𝑥𝑥1 − 𝑥𝑥2 |
|𝑍𝑍| =
𝜎𝜎12 𝜎𝜎22
�
𝑛𝑛1 + 𝑛𝑛2
16.7 − 16.4
Under H0 Z = = 2.71
(0.5) 2 (0.6) 2
+
40 60
Ztab=1.96.
Since Zcal > Ztab, we reject H0 at 5% l.o.s i.e there is a significant difference in the mean length
of fibre manufactured by process A & process B.
5. Pass experience shows that among borewells dug by a firm, 78% are successful. The firm digs
65 borewells in a district. Among them 58 were successful. Can we conclude that these figures
agree with the past experience? [Test both at 5% and 1% levels]
Solution:
H0 : P= 0.78
H1: P≠0.78
p − P0
Z=
p 0 Q0
n
58
− 0.78
Under H0 Z = 65 = 0.219
0.78 × 0.22
65
Ztab=1.96 at 5% l.o.s
Ztab=2.58 at 1% l.o.s
Since Zcal < Ztab, we accept H0 at both the levels. i.e figure agree with past experience at 5% and
1% l.o.s.
6. In a random selection of 55 workers of a factory, 18 were unmarried. Can we conclude that
20% workers of the factory are unmarried?
Solution:
H0 : P= 0.20
H1: P≠ 0.20
p − P0
Z=
p 0 Q0
n
18
− 0.20
Under H0 Z = 55 = 0.271
0.2 × 0.8
55
Ztab=1.96 at 5% l.o.s
Since Zcal < Ztab at 5%, l.o.s, we accept H0 i.e. 20% of the workers in the factory are unmarried.
7. Among 80 electric bulbs manufactured by Process A, (7) were defective. Among 130 electric
bulbs manufactured by process B, two were defective. Test whether the proportion of defectives
in the two process differ.
Solution:
H0 : P1= P2
H1: P1≠ P2
7 2
−
Under H0 Z = 80 130 = 2.506
9 201 1 1
+
210 210 80 130
Ztab=1.96
Since Zcal > Ztab at 5%, l.o.s, we rejected H0 i.e. 5% l.o.s i.e proportion of defective bulbs differ
in process A & process B.
8. The proportion of substandard crackers among 400 crackers manufactured by firm A is 0.12.
The proportion among 500 crackers manufactured by firm B is 0.08. Test at 1% level of
significance that the proportion is the same among the products of the two firms
Solution:
H0 : P1= P2
H1: P1≠ P2
0.12 − 0.08
Under H0 Z = = 1.988
(0.1)(0.9) 1 + 1
400 500
Ztab=2.58 at 1% l.o.s
Since Zcal <Ztab, we accepts H0 at 1% l.o.s. i.e proportion is same among the products of the two
firms.
