Chapter 3 hypothesis testing

CHAPTER 3
3. Hypothesis testing
3.1 Basic concepts
Hypothesis testing is one way of making statistical inference. Hypothesis testing helps us
to determine whether a hypothesis is strongly supported by information from sample data
or not.

Definition:

 Statistical hypothesis is a statement or assumption about population parameter whose

validity is to be checked using statistical method.

 Test statistic: is a statistics whose value serves to determine whether to reject or accept
the hypothesis to be tested. It is a random variable.

 Statistic test: is a test or procedure used to evaluate a statistical hypothesis and its value
depends on sample data.

There are two types of hypothesis:

Null hypothesis:
- It is the hypothesis to be tested.
- It is the hypothesis of equality or the hypothesis of no difference.
- Usually denoted by H0.
Alternative hypothesis:
- It is the hypothesis available when the null hypothesis has to be rejected.
- It is the hypothesis of difference.
- We use non-equality such us: ,  and  signs in alternative hypothesis.
- Usually denoted by H1 or Ha
Example: A language instructor believes that the average age of students in this class is
20 years. In this case, the null and alternative hypothesizes are give as follow:
H 0 : The average age of students in this class is 20 years.

H 1 : Not H 0 . Symbolically

Page 1 of 7
 Chapter 3 hypothesis testing

H 0 :  = 20
H 1 :   20 which is calledTwo sided hypothesis
H 1 :   20,   20 which are called one sided hypothesis

Based on information contained in sample data we can either reject or not null
hypothesis. A choice of rejecting or not of H 0 is based on hypothesis testing.

Important concepts in testing statistical hypothesis

Statistic whose value serves whether to reject H 0 or not is called test statistic. A set of

values which lead us to reject H 0 is called rejection region or critical region. A boarder

line between rejection region and acceptance region is called critical value. The area of
rejection region denoted by  is called level of significance. In any decision making there
are four types of decision.
Reality
H 0 is true H 0 is false

Reject H 0 Wrong (type I error) Correct

Do not Reject H 0 Correct Wrong (type II error)

Decision

Types of error in hypothesis testing

i. Type I error: is error committed if we reject a true null hypothesis.
ii. Type II error: is error committed if we fail to reject a false null hypothesis.
➢ P(type I error ) =  (level of significan ce )

➢ P(type II error ) = 

We cannot minimize  and  simultaneously. Because as  increase  decrease and

vice versa. Since type I error is more serious, we fix  at some smaller value and look for
a test that minimize  . Such type of test is called the most power full test. Power of
test=1-  .
Hypothesis testing for one population mean
To test hypothesis about one population, the different types of the hypotheses are:
i. H 0 :  = 0 versus H1 :   0

Page 2 of 7
 Chapter 3 hypothesis testing

ii. H 0 :  = 0 versus H1 :   0

iii. H 0 :  = 0 versus H1 :    0

Hypothesis testing about  when  2 is known

Here the test statistic is x . If the sample is from normal population with known  2 or if
sample size is large, then the sampling distribution of x is normal.
2
Under H 0 , x ~ N (  o , ) . Consider H 0 :  = 0 versus H1 :   0 . In this case, we
n
reject H 0 if Z  Z  2 or Z  − Z  2  Z  Z  2

Here Z is the test statistics and Z  2 is the critical value.

Similarly we have:
ii. For H 0 :  = 0 versus H1 :   0

x − 0
if Z cal =  Z
 n

iii . For H 0 :  =  0 versus H1 :   0

x − 0
if Z cal =  −Z
 n
Summary of decision making if the population is normal with known variance or if
sample size (n) is large:
No Alternative Hypothesis Rejected H 0 if Do not rejected H 0 if

1 H1 :    0 Z cal  Z 2 Z cal Z 2

2 H1 :    0 Z cal  Z  Z cal  Z 

3 H1 :    0 Z cal  −Z  Z cal  −Z 

Note: If  2 is unknown, but n is large, use s 2 in place of  2 use the above procedure.
The general steps in hypothesis testing are:
Step1. Formulate the appropriate null hypothesis and alternative
Step2.Determine the level of significance

Page 3 of 7
 Chapter 3 hypothesis testing

Step3.Compute the test statistic

Step4.Identify the appropriate critical value
Step5. Identify the rejection region (state the rejection rule)
Step6.Make your decision
Step7.Draw your conclusion
Example: The average score on a nationally administered exam was 28 with standard
deviation of 5. In order to evaluate education system of particular region a sample of 40
students was randomly selected from that region and it was found that, the sample mean
score was 29.8. Can we conclude that there is no difference between the average score of
that region and national students at 5% level of significance assume the score of the
students is normally distributed?
Solution: Given  0 = 28,  = 5, n = 40, x = 29.8,  = 0.05 and population is normal.

Step1. H 0 :  = 28 versus H 1 :   28

Step2.  = 0.05
x − 0 29.8 − 28
Step3. Z cal = = = 2.28
 n 5 40

Step4. The critical value is Z  2 = Z 0.05 2 = Z 0.025 = 1.96

Step5. Reject H 0 if Z cal  Z 2

Step6. Since Z cal = 2.28  Z  2 = 1.96 , we reject H 0

Step7. There is evidence showing that there is significant difference between the national
average score for that of the selected region
Exercise: Do the above problem for 1% level of significance.
Example: A normal population has all unknown mean and variance 13. Ten observations
are chosen at random from this population has mean 12.4. Test whether the population
mean is exceeds 10 at 5% level of significance.
Solution: Given that population is normal,  0 = 10,  2 = 13, n = 10, x = 12.4,  = 0.05

Step1. H 0 :  = 10 versus H 1 :   10

Step2.  = 0.05

Page 4 of 7
 Chapter 3 hypothesis testing

x − 0 12.4 − 10
Step3. Z cal = = = 2.1
 n 13 10

Step4. The critical value is Z  = Z 0.05 = 1.645

Step5. Reject H 0 if Z cal  Z 

Step6. Since Z cal = 2.1  Z  = 1.645 , we reject H 0

Step7. There is sufficient evidence showing that the population mean exceeds 10.
Hypothesis testing about mean when  2 unknown and n-small
If a sample of small size (n) taken from normally distributed population with unknown
x − 0
variance  2 , we estimate  2 by s 2 , and t cal = ~ t (n −1) . In this case the critical value
s n

will be t (n −1) or t 2(n −1) .nye

Summary of decision making if the population is normal with unknown variance and
sample size (n) is small:
No Alternative Hypothesis Rejected if H 0 Do not rejected H 0 if

1 H1 :    0 t cal  t 2 (n − 1) t cal t 2 (n − 1)

2 H1 :    0 t cal  t (n − 1) t cal  t (n − 1)

3 H1 :    0 t cal  −t (n − 1) t cal  −t (n − 1)

Example: A sample of 15 observations taken from normal population which has sample
mean of 38.8 and standard deviation of 2.56. Test whether population mean is 40 or not
at 5% level of significance.
Solution: Given  0 = 40, s = 2.56, n = 15, x = 38.8,  = 0.05 and population is

normal.
Step1. H 0 :  = 40 versus H 1 :   40

Step2.  = 0.05
x − 0 38.8 − 40
Step3. t cal = = = −1.82
s n 2.56 15

Step4. The critical value is t 2 (n − 1) = t 0.05 2 (14) = t 0.025 (14) = 2.145

Page 5 of 7
 Chapter 3 hypothesis testing

Step5. Reject H 0 if t cal  t 2 (n − 1)

Step6. Since t cal = 1.82  t 2 (n − 1) = 2.145 , we do not reject H 0

Step7. Population mean is not significantly different from 40.

Exercise: The present best selling remedy for head-ache is reported bring relief in 15
minutes. Scientist at pharmaceutical company found that a randomly selected 12
individuals their new drug brought relief in an average of 13.5 minute with standard
deviation of 1.5 minute. At  = 0.01 level of significance, is there a reason to believe
that the average relief time for the new medication is shorter than that for old? Assume
the population is normally distributed.
Hypothesis testing about the population proportion
Since the sampling distribution of sample proportion P̂ is normal for large sample size,
then the hypothesis testing for population proportion is similar to that of population
mean. If the hypothesized value of proportion is P0 , then, the hypothesis may be either of:

i. H 0 : P = P0 versus H1 : P  P0

ii. H 0 : P = P0 versus H1 : P  P0

iii. H 0 : P = P0 versus H1 : P  P0

Decision rule
No Alternative Hypothesis Rejected if H 0 Do not rejected H 0 if

1 H 1 : P  P0 Z cal  Z 2 Z cal  Z 2

2 H 1 : P  P0 Z cal  Z  Z cal  Z 

3 H 1 : P  P0 Z cal  −Z  Z cal  −Z 

Pˆ − P0
Z cal = ~ N (0, 1)
P0 (1 − P0 )
n
Example: Among from students in RVU Bishoftu campus, a random sample 250
students was taken and out of these students 38 are found to be left handed. Can it be
concluded that the proportion of left handed students in RVU Bishoftu campus is 0.12?
Use  = 0.02

Page 6 of 7
 Chapter 3 hypothesis testing

Solution: Given P0 = 0.12, Pˆ = 38 250 = 0.152, n = 250,  = 0.02 and population is

normal.
Step1. H 0 : P = 0.12 versus H 1 : P  0.12

Step2.  = 0.02

Pˆ − P0 0.152 − 0.12
Step3. Z cal = = = 1.56
P0 (1 − P0 ) 0.12(1 − 0.12)
n 250
Step4. The critical value is Z  2 = Z 0.02 2 = Z 0.01 = 2.325

Step5. Reject H 0 if Z cal  Z 2

Step6. Since Z cal = 1.56  Z 2 = 2.325 , we do not reject H 0

Step7. There is evidence supporting that the proportion of left handed students in RVU
Bishoftu campus is not significantly different from 0.12.

Page 7 of 7