APPLIED STATISTICS

ECN 416
Theory of Estimation and Test of
Hypothesis Part 2
 TEST OF HYPOTHESIS

• A statistical hypothesis is a statement about

the forms and parameters of the probability
distribution of a random variable.
 TEST OF HYPOTHESIS Contd.
Example:
• Consider a random variable X. The statement
that X is normally distributed, i.e. X ~N(µ, σ2) is
a statistical hypothesis.
• The statement that the average height of the
students in this class is 1.6 m is also a
statistical hypothesis.
 Simple versus Composite Hypothesis

• A hypothesis which is equivalent to a

complete specification of the distribution of a
random variable is called a simple hypothesis.
• Otherwise it is called a composite hypothesis.
• Note: A distribution is completely specified by
giving the functional form and all the
parameters of the distribution.
 Simple versus Composite Hypothesis

• If a population is known to be normal with a

variance σ2, the hypothesis that µ = µ0 is a
simple hypothesis since the mean and the
variance both specifies a normal distribution
completely.
• If however, the variance is unknown, the
hypothesis is said to be composite hypothesis.
  Test of Hypothesis

• A test of hypothesis is a procedure for

deciding whether to accept a statistical
hypothesis i.e. determining whether or not
the hypothesis is compatible with the
available data.
 Some Underlying Assumptions of Test of Hypothesis

• Independence: That is the observations are

independent.
• Randomness: That is the samples are randomly
selected.
• Normality: The population of the observations is
assumed to be normally distributed.
• Homogeneity of the Variances: This assumption
is important in the case of two or more
populations.
 The Null Hypothesis
• The Null hypothesis is the hypothesis being
tested in a test of hypothesis. It is denoted by
H0.
• The null hypothesis is a hypothesis of no
changes i.e. no difference.
• It is often an allegation of the statement or
claim of practical interest actually being
tested.
 The Null Hypothesis Contd

These are three possibilities:

• H0 : µ = µ0
• H0 : π = π
• H0 : σ21 = σ22
 The Alternative Hypothesis
• The hypothesis, statement, or claim of
practical interest is called the alternative
hypothesis denoted by H1.
• It is the hypothesis that there are some
changes or differences between two or more
given situations or that a state of affairs exists.
 The Alternative Hypothesis Contd.
• The possibilities for H1 are:
• H1 : µ ǂ µ 0
• H1 : µ ˂ µ 0
• H1 : µ ˃ µ 0
•  
• H1 : σ21 ǂ σ22
• H1 : σ21 ˂ σ22
• H1 : σ21 ˃ σ22
 The Alternative Hypothesis Contd.
• The alternative hypothesis may be directional or
non- directional.
• When H1 only asserts that the population
parameter is different from the one hypothesized,
it is referred to as non-directional or two-tailed
hypothesis. In this case, the direction or sense of
the change or difference is not specified. A test of
such two-tailed hypothesis is called a two-tailed
or two-sided test.
The Alternative Hypothesis Contd.
Example
• A test of the form
• H1 : µ ǂ µ0
• H1 : µ1 ǂ µ2
• For such test, the critical region is on both
sides of the sampling distribution of the test
statistics.
• (Show the sampling distribution)
The Alternative Hypothesis Contd.
• Occasionally, the alternative hypothesis may
be directional. That is, it takes the direction
and sense of the change or difference.
• Such a hypothesis is called a one-sided or one-
tailed hypothesis and the corresponding test is
called a one-tailed or one-sided Test.
 The Alternative Hypothesis Contd.
The possibilities are listed below:
• H1 : µ ˃ µ 0
• H1 : µ ˂ µ 0
• H1 : σ 2 ˂ σ 20
• H1 : σ 2 ˃ σ 20
•  
• For a one-tailed test the critical region is the region on
only one side or tail of the sampling distribution of the
test statistic.
 Decision
On the basis of the sample we take various
possible decisions or actions namely:
• Reject H0
• Accept H0
• Take further observations (in case of non-fixed
sample sizes e.g., double sampling and
sequential sampling)
 Test Statistic

• A test statistic is a numeral value computed

from the sample observations on the
assumption that the population follows a
particular distribution (e.g., Binomial, Poisson,
Normal, etc. ) used to decide whether or not
to reject the null hypothesis H0 .
 Test Statistic

• In the case of hypothesis concerning the

population mean µ , the test statistic used are:
Z = ( x - µ) / (σ / √n) ~ N(0,1)
and
t = ( x - µ) / (s / √n) ~ t(n-1)
 Critical Region versus Acceptance Region

• On the assumption that the null hypothesis is true, the

appropriate sampling distribution is defined for the test
statistic.
• This sampling distribution is partitioned into two regions
namely the rejection region called the critical region and
the acceptance region which is its complement.
• The critical region of a test is that region such that the
null hypothesis is rejected if the test statistic computed
form a particular sample of a given size takes on a value
that lie within the region.
 Critical Region versus Acceptance Region
Contd.

• The problem of hypothesis testing is one of

determining the best way of partitioning the
sampling distribution of the statistic into the
critical region and the acceptance region.
 Type 1 Error

• This is the error committed by rejecting the

null hypothesis H0 when it is true.
• The probability of committing Type 1 error is
called the level of significance or size of the
test, and it is denoted by α.
α = P (Reject H0 | H0 is True)
 Type 2 Error
• This is the error committed by accepting H0
when it is false. The probability of committing
type 2 error is denoted by β.

β = P (Accepting H0 | H0 is False)

• The Decision Table

 The Operating Characteristic Function of a
Test
• The operating characteristic function of a test is a
function denoted by β(Ө) showing the
relationship between the probability of accepting
H0 and the various possible values of the
parameter as specified by H1.
• The graph β(Ө) against Ө is called the operating
characteristic curve of the test.
• They provide the picture of the various risk to
which we are exposed in a test hypothesis.
 Power of a Test

• The power of a test is defined as the

probability of rejecting H0 when it is false. It is
the probability of making the correct decision
and it is denoted by 1- β.

1- β = P (Rejecting H0 | H1 is True)
 Power Curve
• A graph of 1- β against various alternative
values of Ө is called the Power Curve.
• The higher the power of the test the stronger
its discriminating power between a true and
false hypothesis.
•  If there exists a test that is uniformly more
powerful than any other test of the same size α,
we call such test the Uniformly Most Powerful
Test.
 BASIC STEPS IN HYPOTHESIS TESTING

• State the underlying assumptions

• State H0 and H1
• State the level of significance α
• State the test statistic and compute the
observed value of the test statistic.
• State the decision rule
• State the conclusions.
 SOME SPECIFIC TESTS

A. Testing for µ when σ2 is known, the

population is Normal or n ˃ 30 for a non-
Normal population
 SOME SPECIFIC TESTS Contd.

H0 H1 Reject H0 , If :
µ = µ0 µ ǂ µ0 |Z| > Z 1-α/2
µ = µ0 µ > µ0 Z > Z 1-α
µ = µ0 µ < µ0 Z < - Z 1-α

Z = ( x - µ) / (σ / √n) ~ N(0,1)
 SOME SPECIFIC TESTS Contd.
Example
From a normal population of variance 16 units, a
random sample of 36 items gave a mean value
of 84 units at the 5% level of significance. Test
the hypothesis:
H1 : µ ǂ 80
H1 : µ > 80
 SOME SPECIFIC TESTS Contd.
Solution
(a) H0 : µ = 80 H1 : µ ǂ 80 α = 0.05
Z = (84 - 80) / (4 / √36) = 5.97

Z 1-α/2 = Z 0.975 = 1.96

Conclusion: Since |5.97| > 1.96, Reject H0. The

data support the claim that µ ǂ 80
 SOME SPECIFIC TESTS Contd.
(b) H0 : µ = 80 H1 : µ > 80 α = 0.05
Z = (84 - 80) / (4 / √36) = 5.97
Z 1-α = Z 0.95 = 1.645

Conclusion: Since 5.97 > 1.645, Reject H0.

Therefore the claim that µ > 80 is justified
by the available data.
 SOME SPECIFIC TESTS Contd.

B. Testing for µ when σ2 is unknown, and or

the population is Normal with n ≤ 30
 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :
µ = µ0 µ ǂ µ0 |t| > t α/2 (n-1)
µ = µ0 µ > µ0 t > t α (n-1)
µ = µ0 µ < µ0 t < - t α (n-1)

t = ( x - µ) / (s / √n) ~ t(n-1)
 SOME SPECIFIC TESTS Contd.
Example
In an intelligence test on Ten Students, the
following scores were obtained :
105, 120, 90, 65, 130, 110, 120, 115, 125, 100.
Given that the average score for the class before a
specified tutorial for the test was 105, has the
special tutorial improved the performance of the
students? Let α = 1%
 SOME SPECIFIC TESTS Contd.
Solution
H0 : µ = 105H1 : µ > 105 α = 0.01,
x = 110s = 8
t = (110 - 105) / (8 / √ 10) = 1.96
t α (n-1) = t 0.01 (9) = 2.82

Conclusion: Since 1.96 < 2.82, therefore we accept H0.

There is no evidence to show that the specified tuition
have improved the students’ performance.
SOME SPECIFIC TESTS Contd.

C. Testing for π
 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :
π = π0 π ǂ π0 |Z| > Z 1-α/2
π = π0 π > π0 Z > Z 1-α
π = π0 π < π0 Z < - Z 1-α
 
Z = ( p - π) / σp
= ( p - π) / sp ~ N(0,1)
 SOME SPECIFIC TESTS Contd.
Example
A political party stalwart claims that he has the
support of 55% of the voters in his constituency.
What will the party executive committee
conclude if out of a random sample of 500
registered voters only 245 express their
preference for him? Take α = 0.05.
 SOME SPECIFIC TESTS Contd.
Solution
H0 : π = 0.55 H1 : π < 0.55 α = 0.05,
p = √ [(0.49)(0.51) / 500] = 0.022
Z = (0.49 – 0.55) / 0.022 = - 2.73
-Z 1- α = -Z 0.95 = -1.64
 
Conclusion: Since -2.73 < -1.64, therefore we reject H0 .
Therefore the claim by the party stalwart of the support of
55% of the voters in his constituency is not justified by the
available data.
 SOME SPECIFIC TESTS Contd.

D. Test for µ1 - µ2.

• When σ12, σ22 are known, the populations are
Normal or n1 , n2 ˃ 30 for a non-Normal
populations
 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :
µ1 = µ 2 µ1 ǂ µ 2 Z| > Z 1-α/2
µ1 = µ 2 µ1 > µ 2 Z > Z 1-α
µ1 = µ 2 µ1 < µ 2 Z < - Z 1-α
 
Z = [( x1 - x2 ) - (µ1 - µ2 )] / (σ x1 – x2) ~ N(0,1)
 SOME SPECIFIC TESTS Contd.
Remark:
Provided n1 , n2 ˃ 30, the statistic
Z = [( x1 - x2 ) - (µ1 - µ2 )] / (σ x1 – x2) ~ N(0,1)
Where
σ x1 – x2 = √[ (s12 / n1) + (s22 / n2) ]
this can be used in cases where σ12, σ22 are
unknown
 SOME SPECIFIC TESTS Contd.
Example
• A random sample of 100 students attending
Lagos State University have a mean age of 23
years and a standard deviation of 4 years while a
random sample of 50m students from University
of Lagos have a mean age of 21 years and a
standard deviation of 5 years. Can we conclude
that students in Lagos State University are older
than students in University of Lagos? Take α = 5%.
 SOME SPECIFIC TESTS Contd.
Solution
n1 = 100, x1 = 23, s1 = 4 : n2 = 50, x2 = 21, s2 = 5
H0: µ1 - µ2 H1: µ1 > µ2 α = 0.05,
σ x1 – x2 = √ [ (42 / 100) + (52 / 50)] = 0.812
Z = [ (23 – 21) - 0 ] / 0.812 = 2.463
Z 1-α = Z 0.95 = 1.646

Conclusion: Since 2.463 > 1.646, Reject H0.

There is evidence to show that students in Lagos State University
are older than Students in University of Lagos.
 SOME SPECIFIC TESTS Contd.

D. Test for µ1 - µ2.

• When σ12, σ22 are unknown, the populations
are Normal but n1 , n2 < 30 for a non-Normal
populations.
i. Assuming σ12 = σ22 = σ2
 SOME SPECIFIC TESTS Contd.

H0 H1 Reject H0 , If :
µ1 = µ2 µ1 ǂ µ2 | t | > t α/2 (v) df
µ1 = µ 2 µ1 < µ 2 t > t α (v) df
µ1 = µ 2 µ1 > µ 2 t < - t α (v) df
 
 SOME SPECIFIC TESTS Contd.
where v = n1 + n2 -2

t = [(x1 - x2 ) - (µ1 - µ2 )] / ( s x1 – x2 ) ~ t(v) df

s x1 – x2 = √ [sp2 (1/ n1 + 1/ n2)]

sp2 = [(n1 - 1) s1 + (n2 - 1) s2 ] / n1 + n2 -2
 SOME SPECIFIC TESTS Contd.
Example
An aptitude test was given to a random sample of 12 science
students and 10 Arts students selected from a class of final year
students of Faith Academy.

The result are as follows:

Group 1: x1 = 74, s1 = 8; Group 2: x2 = 70, s2 = 10

Assuming that the variances of the two groups are the same, is
there any significance difference between the means of the two
groups at 5% level?
 SOME SPECIFIC TESTS Contd.
Solution

Assuming σ12 = σ22 = σ2

H0: µ1 - µ2 H1: µ1 ǂ µ2 α = 0.05,
sp2 = [(12 - 1) 82 + (10 - 1) 102] / (8 + 10 - 2) = 80.2
sp = 8.96
 
s x1 – x2 = 8.96 √ (1/12 + 1/10) = 3.85
 
t = [(74 – 70) - 0 ] / 3.85 = 1.038
t0.025 (20) = 2.09
 

Conclusion: Since | 1.038 | < 2.09, Accept H0. There is nothing to show that the difference
between the means of the two groups of students is significant at 5% level.
 SOME SPECIFIC TESTS Contd.

D. Test for µ1 - µ2.

• When σ12, σ22 are unknown, the populations
are Normal but n1 , n2 < 30 for a non-Normal
populations.
ii. Assuming σ12 ǂ σ22
 SOME SPECIFIC TESTS Contd.

H0 H1 Reject H0 , If :
µ1 = µ2 µ1 ǂ µ2 | t | > t α/2 (v/) df
µ1 = µ 2 µ1 > µ 2 t > t α (v/) df
µ1 = µ 2 µ1 < µ 2 t < - t α (v/) df
 SOME SPECIFIC TESTS Contd.
Where
σ x1 – x2 = √ [ (s12 / n1) + (s22 / n2) ]
and
v/ = [(s12 / n1) + (s22 / n2)]2 / [(s12 / n1)2 / n1 +
(s22 / n2)2 / n2]
 SOME SPECIFIC TESTS Contd.

Example

Recall the data in the last example assuming that

the variances of the two groups are different. Is
the difference between the means of the two
groups significant at 5% level?
 SOME SPECIFIC TESTS Contd.
Solution
Assuming σ12 ǂ σ22
H0: µ1 - µ2 H1: µ1 ǂ µ2 α = 0.05,
σ x1 – x2 = √ [(82 / 12) + (102 / 10)] = 3.92

v/ = [(82 / 12 ) + (102 /10)]2 / {[ (82 / 12)2 / 12 ] + [(102 /10)2 / 10]}

= 18.99 = 19
t = [(74 – 70) - 0 ] / 3.92 = 1.02
t(v/ ) = t0.025 (19) = 2.09
 
Conclusion: Since | 1.02 | < 2.09, accept H0 Therefore, the data does not support
the claim that there is nay significant difference between the two groups of
students.
SOME SPECIFIC TESTS Contd.

E. Test for π1 - π2.

 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :
π 1 = π 2 π 1 ǂ π2 |Z| > Z 1-α/2
π1 = π2 π1 > π2 Z > Z 1-α
π1 = π2 π1 < π2 ` Z < - Z 1-α

Z = [( p1 - p1) - ( π1 – π2 )] / σp1 – p2 ~ N(0,1)

 SOME SPECIFIC TESTS Contd.
Example
200 patients suffering from a certain disease were
randomly divided into two equal groups. Of the
first group who received the standard treatment,
78 recovered within 3 days. Out of the 100 who
were treated by a new method, 90 recovered
within 3 days. First, the doctor claimed that the
standard treatment is less effective than the new
treatment. Is this claim correct? Take α = 0.05.
 SOME SPECIFIC TESTS Contd.
Solution
H0: π1 = π2 H1: π1 < π2 α = 0.05.
Sp1 – p2 = √ {[(0.78)(0.22) / 100] + [(0.90)(0.10) / 100]} =
0.051
Z = (0.78 - 0.90) / 0.051 = -2.35
Z1 – α Z0.95 1.64
 
Conclusion: Since -2.35 < -1.64, Reject H 0 . The data support
the claim that the standard treatment is less effective than the
new treatment.
SOME SPECIFIC TESTS Contd.

F. Test for σ2.

 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :
σ2 = σ02 σ2 ǂ σ02 χ2 < χ21-α/2 (n – 1)
OR χ2 > χ2α/2 (n – 1)
σ2 = σ02 σ2 > σ02 χ2 > χ2α (n – 1)
σ2 = σ02 σ2 < σ02 ` χ2 < - χ2α (n – 1)
Where
χ2 = (n – 1) s2 / σ02 ~ χ2 (n – 1)
 SOME SPECIFIC TESTS Contd.
Example
Consider a sample of 15 items. The sample
variance s2 = 1225. Test the hypothesis
σ2 = 2500 against σ2 ǂ 2500. Use α = 0.05
σ2 = 2500 against σ2 > 2500. Use α = 0.01
 
 SOME SPECIFIC TESTS Contd.
Solution
H0 : σ2 = 2500 H1 σ2 ǂ 2500. α = 0.05
χ2 = (14)(1225) / 2500 = 6.86
χ20.975 (14) = 5.63
χ20.025 (14) = 26.12
Reject Ho unless χ21-α/2 (n – 1) ≤ χ2 ≤ χ2α/2 (n – 1)
 
Conclusion: Since 5.63 ≤ 6.86, Accept Ho. There is no
evidence to show that σ2 is different from 2500.
 SOME SPECIFIC TESTS Contd.
H0 : σ2 = 2500 H1 σ2 > 2500. α = 0.01
χ2 = (14)(1225) / 2500 = 6.86
χ20.01 (14) = 29.14
 
Conclusion: Since 6.86 < 29.14. Accept H0. The
data does not support the claim that σ2 = 2500
SOME SPECIFIC TESTS Contd.

F. Test for σ2.

 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :
σ2 = σ02 σ2 ǂ σ02 χ2 < χ21-α/2 (n – 1)
OR χ2 > χ2α/2 (n – 1)
σ2 = σ02 σ2 > σ02 χ2 > χ2α (n – 1)
σ2 = σ02 σ2 < σ02 ` χ2 < - χ2α (n – 1)
Where
χ2 = (n – 1) s2 / σ02 ~ χ2 (n – 1)
 SOME SPECIFIC TESTS Contd.
Example
Consider a sample of 15 items. The sample
variance s2 = 1225. Test the hypothesis
σ2 = 2500 against σ2 ǂ 2500. Use α = 0.05
σ2 = 2500 against σ2 > 2500. Use α = 0.01
 SOME SPECIFIC TESTS Contd.
Solution
H0 : σ2 = 2500 H1 σ2 ǂ 2500. α = 0.05
χ2 = (14)(1225) / 2500 = 6.86
χ20.975 (14) = 5.63
χ20.025 (14) = 26.12
Reject Ho unless χ21-α/2 (n – 1) ≤ χ2 ≤ χ2α/2 (n – 1)

Conclusion: Since 5.63 ≤ 6.86, Accept Ho

There is no evidence to show that σ2 is different from 2500.
 SOME SPECIFIC TESTS Contd.
H0 : σ2 = 2500 H1 σ2 > 2500. α = 0.01
χ2 = (14)(1225) / 2500 = 6.86
χ20.01 (14) = 29.14
 
Conclusion: Since 6.86 < 29.14. Accept H0
The data does not support the claim that σ2 =
2500
 SOME SPECIFIC TESTS Contd.

G. Test for σ12/ σ22 = 1, σ12 = σ22.

 SOME SPECIFIC TESTS Contd.
H0 H1 Reject H0 , If :

σ12 = σ22 σ12 ǂ σ22 F < F1-α/2 (v1, v2)

OR F < Fα/2 (v1, v2)
σ12 = σ22 σ12 > σ22 F > F1-α (v1, v2)
σ12 = σ22 σ12 < σ22 F < F1-α (v1, v2)
 SOME SPECIFIC TESTS Contd.

Example
Consider 2 random samples of sizes 11, 25. If the
sample variances are 35.31 meters and 15.67
meters respectively, test the hypothesis that
σ12 = σ22 against the alternatives
σ12 ǂ σ22 at 5% level
σ12 < σ22 at 1% level
 SOME SPECIFIC TESTS Contd.
 H0 : σ12 = σ22 H1: σ12 ǂ σ22 α = 0.05 

F = s12 / s22 = 35.31 / 15.67 = 2.25

F0.025 (10, 24) = 2.64
F0.975 (10, 24) = 1 / F0.025 (24, 10)
= 1 / 3.37 = 0.297
 
Conclusion: Since 0.297 ≤ 2.25 ≤ 2.64, Accept H0
 SOME SPECIFIC TESTS Contd.
H0 : σ12 = σ22 H1: σ12 < σ22 α = 0.01
 
F = s12 / s22 = 35.31 / 15.67 = 2.25
F0.90 (10, 24) = 1 / F0.10 (24, 10)
= 1 / 2.18 = 0.459

Conclusion: Since 2.25 > 0.459. Accept H0.

Therefore the available data support the claim that the two
samples were drawn from populations with equal variances.