Chapter 6.

Hypothesis Testing

Nguyễn Văn Hạnh

Department of Applied Mathematics
School of Applied Mathematics and Informatics

Second semester, 2022-2023

NV HANH Probability and statistics Second semester, 2022-2023 1 / 50

Contents

1 Introduction to hypothesis testing

2 Test on a population mean µ.

3 Test on a population proportion p.

4 Test on the difference between two means µ1 − µ2 .

5 Test on the difference between two proportions p1 − p2

NV HANH Probability and statistics Second semester, 2022-2023 2 / 50

 Introduction to hypothesis testing

Introduction

Example 6.1:
Past experience indicates that a monthly long-distance telephone bill
is normally distributed with µ0 = 17.85 (USD) and a standard
deviation σ = 3.39.

NV HANH Probability and statistics Second semester, 2022-2023 3 / 50

 Introduction to hypothesis testing

Introduction

Example 6.1:
Past experience indicates that a monthly long-distance telephone bill
is normally distributed with µ0 = 17.85 (USD) and a standard
deviation σ = 3.39.
After an advertising campaign aimed at increasing long-distance
telephone usage, the manager of the telephone company took a
random sample of 25 household bills and recorded the mean of their
monthly usage x̄ = 19.93.

NV HANH Probability and statistics Second semester, 2022-2023 3 / 50

 Introduction to hypothesis testing

Introduction

Example 6.1:
Past experience indicates that a monthly long-distance telephone bill
is normally distributed with µ0 = 17.85 (USD) and a standard
deviation σ = 3.39.
After an advertising campaign aimed at increasing long-distance
telephone usage, the manager of the telephone company took a
random sample of 25 household bills and recorded the mean of their
monthly usage x̄ = 19.93.
Do the data allow us to infer that the campaign was successful?

NV HANH Probability and statistics Second semester, 2022-2023 3 / 50

 Introduction to hypothesis testing

Introduction

Example 6.1:
Past experience indicates that a monthly long-distance telephone bill
is normally distributed with µ0 = 17.85 (USD) and a standard
deviation σ = 3.39.
After an advertising campaign aimed at increasing long-distance
telephone usage, the manager of the telephone company took a
random sample of 25 household bills and recorded the mean of their
monthly usage x̄ = 19.93.
Do the data allow us to infer that the campaign was successful?
→ We need to make a decision of accepting or rejecting the hypothesis
that the campaign was successful (µ > 17.85, the mean monthly bill µ
after the advertising campaign is greater than 17.85).

NV HANH Probability and statistics Second semester, 2022-2023 3 / 50

 Introduction to hypothesis testing

Introduction
Example 6.2: It is known that the mean monthly electricity bills of
households in a region in March 2022 is 225.5 (thousands dong).
Observed the monnthly electricity bills of 200 households in April 2022 and
the data are given below:

NV HANH Probability and statistics Second semester, 2022-2023 4 / 50

 Introduction to hypothesis testing

Introduction

The distribution of data can be approximated by a normal distribution:

NV HANH Probability and statistics Second semester, 2022-2023 5 / 50

 Introduction to hypothesis testing

Introduction

Modelling: We can suppose that the monnthly electricity bills X of

households in this region in April 2022 follow a normal distribution
with parameters θ = (µ, σ 2 ).

NV HANH Probability and statistics Second semester, 2022-2023 6 / 50

 Introduction to hypothesis testing

Introduction

Modelling: We can suppose that the monnthly electricity bills X of

households in this region in April 2022 follow a normal distribution
with parameters θ = (µ, σ 2 ).
The parameter µ is the population mean (the mean monnthly
electricity bills of all households in this region in April 2022).

NV HANH Probability and statistics Second semester, 2022-2023 6 / 50

 Introduction to hypothesis testing

Introduction

Modelling: We can suppose that the monnthly electricity bills X of

households in this region in April 2022 follow a normal distribution
with parameters θ = (µ, σ 2 ).
The parameter µ is the population mean (the mean monnthly
electricity bills of all households in this region in April 2022).
The sample mean x̄ = 236.78 > µ0 = 225.5 ( the mean monnthly
electricity bills of 200 households in April 2022 is greater than that in
March 2022).

NV HANH Probability and statistics Second semester, 2022-2023 6 / 50

 Introduction to hypothesis testing

Introduction

Modelling: We can suppose that the monnthly electricity bills X of

households in this region in April 2022 follow a normal distribution
with parameters θ = (µ, σ 2 ).
The parameter µ is the population mean (the mean monnthly
electricity bills of all households in this region in April 2022).
The sample mean x̄ = 236.78 > µ0 = 225.5 ( the mean monnthly
electricity bills of 200 households in April 2022 is greater than that in
March 2022).
Question: Is there enough evidence to infer that the mean monnthly
electricity bills µ of all households in April 2022 is greater than that in
March 2022?

NV HANH Probability and statistics Second semester, 2022-2023 6 / 50

 Introduction to hypothesis testing

Introduction

Modelling: We can suppose that the monnthly electricity bills X of

households in this region in April 2022 follow a normal distribution
with parameters θ = (µ, σ 2 ).
The parameter µ is the population mean (the mean monnthly
electricity bills of all households in this region in April 2022).
The sample mean x̄ = 236.78 > µ0 = 225.5 ( the mean monnthly
electricity bills of 200 households in April 2022 is greater than that in
March 2022).
Question: Is there enough evidence to infer that the mean monnthly
electricity bills µ of all households in April 2022 is greater than that in
March 2022?
We need to make a decision of accepting or rejecting the hypothesis
”µ > 225.5”.

NV HANH Probability and statistics Second semester, 2022-2023 6 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

Research hypothesis: ”µ > 225.5” is called the research hypothesis:

need to be accepted or rejected.

NV HANH Probability and statistics Second semester, 2022-2023 7 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

Research hypothesis: ”µ > 225.5” is called the research hypothesis:

need to be accepted or rejected.
We state an opposite hypothesis with the research hypothesis and is
called the null hypothesis, labed by H0 : µ ≤ 225.5 (the mean
monnthly electricity bills µ of all households in April 2022 is not
greater than that in March 2022).

NV HANH Probability and statistics Second semester, 2022-2023 7 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

Research hypothesis: ”µ > 225.5” is called the research hypothesis:

need to be accepted or rejected.
We state an opposite hypothesis with the research hypothesis and is
called the null hypothesis, labed by H0 : µ ≤ 225.5 (the mean
monnthly electricity bills µ of all households in April 2022 is not
greater than that in March 2022).
To make a decision of accepting the research hypothesis, we test
whether the information of the sample is strong enough to reject the
null hypothsis H0 .

NV HANH Probability and statistics Second semester, 2022-2023 7 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

Research hypothesis: ”µ > 225.5” is called the research hypothesis:

need to be accepted or rejected.
We state an opposite hypothesis with the research hypothesis and is
called the null hypothesis, labed by H0 : µ ≤ 225.5 (the mean
monnthly electricity bills µ of all households in April 2022 is not
greater than that in March 2022).
To make a decision of accepting the research hypothesis, we test
whether the information of the sample is strong enough to reject the
null hypothsis H0 .
The alternative hypothesis, labeled by H1 : is the research hypothesis
(µ > 2225.5).

NV HANH Probability and statistics Second semester, 2022-2023 7 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

Research hypothesis: ”µ > 225.5” is called the research hypothesis:

need to be accepted or rejected.
We state an opposite hypothesis with the research hypothesis and is
called the null hypothesis, labed by H0 : µ ≤ 225.5 (the mean
monnthly electricity bills µ of all households in April 2022 is not
greater than that in March 2022).
To make a decision of accepting the research hypothesis, we test
whether the information of the sample is strong enough to reject the
null hypothsis H0 .
The alternative hypothesis, labeled by H1 : is the research hypothesis
(µ > 2225.5).
Problem of testing: test the null hypothsis H0 : µ ≤ 225.5 versus the
alternative hypothesis H1 : µ > 225.5.

NV HANH Probability and statistics Second semester, 2022-2023 7 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

04 outcomes of a test:

NV HANH Probability and statistics Second semester, 2022-2023 8 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

04 outcomes of a test:

Type I error: we reject H0 when it is true (we infer that µ > 225.5
but µ is really ≤ 225.5)

NV HANH Probability and statistics Second semester, 2022-2023 8 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

04 outcomes of a test:

Type I error: we reject H0 when it is true (we infer that µ > 225.5
but µ is really ≤ 225.5)
Type II error: we accept H0 when it is false (we infer that µ ≤ 225.5
but µ is really > 225.5)

NV HANH Probability and statistics Second semester, 2022-2023 8 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

P(Type I error) = P(reject H0 | H0 is true) = α.

NV HANH Probability and statistics Second semester, 2022-2023 9 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

P(Type I error) = P(reject H0 | H0 is true) = α.

P(Type II error) = P(accept H0 | H0 is false) = β.

NV HANH Probability and statistics Second semester, 2022-2023 9 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

P(Type I error) = P(reject H0 | H0 is true) = α.

P(Type II error) = P(accept H0 | H0 is false) = β.
Power of test) = 1 − β = P(we accept H1 when it is true).

NV HANH Probability and statistics Second semester, 2022-2023 9 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

P(Type I error) = P(reject H0 | H0 is true) = α.

P(Type II error) = P(accept H0 | H0 is false) = β.
Power of test) = 1 − β = P(we accept H1 when it is true).
If α decreases, then β will increase and and vice versa.

NV HANH Probability and statistics Second semester, 2022-2023 9 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

P(Type I error) = P(reject H0 | H0 is true) = α.

P(Type II error) = P(accept H0 | H0 is false) = β.
Power of test) = 1 − β = P(we accept H1 when it is true).
If α decreases, then β will increase and and vice versa.
We control the P(Type I error), α, at a given threshold, called the
level of significance, and the test is called the test of significance.

NV HANH Probability and statistics Second semester, 2022-2023 9 / 50

 Introduction to hypothesis testing

Hypothesis testing: Definitions

P(Type I error) = P(reject H0 | H0 is true) = α.

P(Type II error) = P(accept H0 | H0 is false) = β.
Power of test) = 1 − β = P(we accept H1 when it is true).
If α decreases, then β will increase and and vice versa.
We control the P(Type I error), α, at a given threshold, called the
level of significance, and the test is called the test of significance.
We usually choose the level of significance α that equals to 1%, 5% or
10%.

NV HANH Probability and statistics Second semester, 2022-2023 9 / 50

 Introduction to hypothesis testing

Procedure of a test

Step 1: State the null hypothesis H0 and the alternaitve hypothesis

H1 .

NV HANH Probability and statistics Second semester, 2022-2023 10 / 50

 Introduction to hypothesis testing

Procedure of a test

Step 1: State the null hypothesis H0 and the alternaitve hypothesis

H1 .
Step 2: Consider the asssumptions of the test.

NV HANH Probability and statistics Second semester, 2022-2023 10 / 50

 Introduction to hypothesis testing

Procedure of a test

Step 1: State the null hypothesis H0 and the alternaitve hypothesis

H1 .
Step 2: Consider the asssumptions of the test.
Step 3: Compute the test - statistic Z : is a statistic that has fully
deterministic distribution when the null hypothesis is true.

NV HANH Probability and statistics Second semester, 2022-2023 10 / 50

 Introduction to hypothesis testing

Procedure of a test

Step 1: State the null hypothesis H0 and the alternaitve hypothesis

H1 .
Step 2: Consider the asssumptions of the test.
Step 3: Compute the test - statistic Z : is a statistic that has fully
deterministic distribution when the null hypothesis is true.
Step 4: Find a rejection region Wα such that
P(Z ∈ Wα |H0 is true ) ≤ α.

NV HANH Probability and statistics Second semester, 2022-2023 10 / 50

 Introduction to hypothesis testing

Procedure of a test

Step 1: State the null hypothesis H0 and the alternaitve hypothesis

H1 .
Step 2: Consider the asssumptions of the test.
Step 3: Compute the test - statistic Z : is a statistic that has fully
deterministic distribution when the null hypothesis is true.
Step 4: Find a rejection region Wα such that
P(Z ∈ Wα |H0 is true ) ≤ α.
Step 5: Make a decision: if the test-statistic Z ∈ Wα then we reject
H0 , otherwise we fail to reject H0 . Give a conclusion about the
research hypothesis.

NV HANH Probability and statistics Second semester, 2022-2023 10 / 50

 Introduction to hypothesis testing

Principle of small probability

Principle of small probability: “Specified events of small probability do

not occur by chance, when the experiment is performed once.”

NV HANH Probability and statistics Second semester, 2022-2023 11 / 50

 Introduction to hypothesis testing

Principle of small probability

Principle of small probability: “Specified events of small probability do

not occur by chance, when the experiment is performed once.”
Decision Rule: if the test-statistic Z ∈ Wα then we reject H0 ,
otherwise we fail to reject H0

NV HANH Probability and statistics Second semester, 2022-2023 11 / 50

 Introduction to hypothesis testing

Principle of small probability

Principle of small probability: “Specified events of small probability do

not occur by chance, when the experiment is performed once.”
Decision Rule: if the test-statistic Z ∈ Wα then we reject H0 ,
otherwise we fail to reject H0 since:
If the test-statistic Z ∈ Wα and we accept H0 , the the event
"Z ∈ Wα |H0 is true " of small probability α has occured by chance.
This contradicts the principle of small probability.

NV HANH Probability and statistics Second semester, 2022-2023 11 / 50

 Introduction to hypothesis testing

Principle of small probability

Principle of small probability: “Specified events of small probability do

not occur by chance, when the experiment is performed once.”
Decision Rule: if the test-statistic Z ∈ Wα then we reject H0 ,
otherwise we fail to reject H0 since:
If the test-statistic Z ∈ Wα and we accept H0 , the the event
"Z ∈ Wα |H0 is true " of small probability α has occured by chance.
This contradicts the principle of small probability.
So if the test-statistic Z ∈ Wα then we reject H0 , otherwise we fail to
reject H0 .

NV HANH Probability and statistics Second semester, 2022-2023 11 / 50

 Introduction to hypothesis testing

Four problems of testing

We study on the following four problems of testing:

Test on a population mean µ.
Test on a population proportion p.
Test on the difference between 2 population means µ1 − µ2 .
Test on the difference between 2 population proportions p1 − p2 .

NV HANH Probability and statistics Second semester, 2022-2023 12 / 50

 Test on a population mean µ.

Test on a population mean µ.

Problem 1: We observe a poppulation X where µ = E (X ) and σ 2 = V (X ).

Consider a random sample (X1 , X2 , ..., Xn ) taken from the population X .

NV HANH Probability and statistics Second semester, 2022-2023 13 / 50

 Test on a population mean µ.

Test on a population mean µ.

Problem 1: We observe a poppulation X where µ = E (X ) and σ 2 = V (X ).

Consider a random sample (X1 , X2 , ..., Xn ) taken from the population X .
We test a research hypothesis on µ that µ > µ0 or µ < µ0 or µ 6= µ0 .

NV HANH Probability and statistics Second semester, 2022-2023 13 / 50

 Test on a population mean µ.

Test on a population mean µ.

Problem 1: We observe a poppulation X where µ = E (X ) and σ 2 = V (X ).

Consider a random sample (X1 , X2 , ..., Xn ) taken from the population X .
We test a research hypothesis on µ that µ > µ0 or µ < µ0 or µ 6= µ0 .
It can be proved that the null hypotheses H0 : µ = µ0 and
H0 : µ ≤ µ0 against the alternative hypothesis H1 : µ > µ0 have the
same rejection region Wα .

NV HANH Probability and statistics Second semester, 2022-2023 13 / 50

 Test on a population mean µ.

Test on a population mean µ.

Problem 1: We observe a poppulation X where µ = E (X ) and σ 2 = V (X ).

Consider a random sample (X1 , X2 , ..., Xn ) taken from the population X .
We test a research hypothesis on µ that µ > µ0 or µ < µ0 or µ 6= µ0 .
It can be proved that the null hypotheses H0 : µ = µ0 and
H0 : µ ≤ µ0 against the alternative hypothesis H1 : µ > µ0 have the
same rejection region Wα .
Thus, we only consider the null hypothesis H0 : µ = µ0 against one of
three types of alternative:
H1 : µ > µ0 : right-tailed test
or H1 : µ < µ0 : left-tailed test
or H1 : µ 6= µ0 : two-tailed test.

NV HANH Probability and statistics Second semester, 2022-2023 13 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

NV HANH Probability and statistics Second semester, 2022-2023 14 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.1: We consider the right-tailed test H1 : µ > µ0 .

NV HANH Probability and statistics Second semester, 2022-2023 14 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.1: We consider the right-tailed test H1 : µ > µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

NV HANH Probability and statistics Second semester, 2022-2023 14 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.1: We consider the right-tailed test H1 : µ > µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

Find the rejection region Wα : We reject H0 and accept H1 if

X̄ >> µ0 or X̄ − µ0 is large enough ⇔ Z = X̄σ/−µ
√ 0 > x, for a critical
n
value x to be found.

NV HANH Probability and statistics Second semester, 2022-2023 14 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.1: We consider the right-tailed test H1 : µ > µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

Find the rejection region Wα : We reject H0 and accept H1 if

X̄ >> µ0 or X̄ − µ0 is large enough ⇔ Z = X̄σ/−µ
√ 0 > x, for a critical
n
value x to be found.
The rejection region Wα = {Z > x} = (x; +∞) satisfies:

P(Type I error) = P(reject H0 |if H0 is true) = P(Z ∈ Wα |H0 is true)

P(Type I error) = P(Z > x|H0 is true) ≤ α ⇔ x = Zα

NV HANH Probability and statistics Second semester, 2022-2023 14 / 50
 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

Case 1.1: We consider the right-tailed test H1 : µ > µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

The rejection region Wα = {Z > Zα } = (Zα ; +∞).

NV HANH Probability and statistics Second semester, 2022-2023 15 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

Case 1.1: We consider the right-tailed test H1 : µ > µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

The rejection region Wα = {Z > Zα } = (Zα ; +∞).

The decision rule: If Z > Zα , then we reject H0 , otherwise we fail to
reject H0 .

NV HANH Probability and statistics Second semester, 2022-2023 15 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.2: We consider the left-tailed test H1 : µ < µ0 .

NV HANH Probability and statistics Second semester, 2022-2023 16 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.2: We consider the left-tailed test H1 : µ < µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

NV HANH Probability and statistics Second semester, 2022-2023 16 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.2: We consider the left-tailed test H1 : µ < µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

Find the rejection region Wα : We reject H0 and accept H1 if

X̄ << µ0 or X̄ − µ0 is small enough ⇔ Z = X̄σ/−µ
√ 0 < x, for a critical
n
value x to be found.

NV HANH Probability and statistics Second semester, 2022-2023 16 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
Case 1.2: We consider the left-tailed test H1 : µ < µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

Find the rejection region Wα : We reject H0 and accept H1 if

X̄ << µ0 or X̄ − µ0 is small enough ⇔ Z = X̄σ/−µ
√ 0 < x, for a critical
n
value x to be found.
The rejection region Wα = {Z < x} = (∞; x) satisfies:

P(Type I error) = P(reject H0 |if H0 is true) = P(Z ∈ Wα |H0 is true)

P(Type I error) = P(Z < x|H0 is true) ≤ α ⇔ x = −Zα

NV HANH Probability and statistics Second semester, 2022-2023 16 / 50
 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

Case 1.2: We consider the left-tailed test H1 : µ < µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

The rejection region Wα = {Z < −Zα } = (−∞; −Zα ).

NV HANH Probability and statistics Second semester, 2022-2023 17 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

Case 1.2: We consider the left-tailed test H1 : µ < µ0 .
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

The rejection region Wα = {Z < −Zα } = (−∞; −Zα ).

The decision rule: If Z < −Zα , then we reject H0 , otherwise we fail
to reject H0 .

NV HANH Probability and statistics Second semester, 2022-2023 17 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
6 µ0 .
Case 1.3: We consider the two-tailed test H1 : µ =

NV HANH Probability and statistics Second semester, 2022-2023 18 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
6 µ0 .
Case 1.3: We consider the two-tailed test H1 : µ =
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

NV HANH Probability and statistics Second semester, 2022-2023 18 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
6 µ0 .
Case 1.3: We consider the two-tailed test H1 : µ =
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

Find the rejection region Wα : We reject H0 and accept H1 if X̄ is

much different from µ0 or |X̄ − µ0 | is large enough
⇔ |Z | = |X̄σ/−µ |
√ 0 > x, for a critical value x to be found.
n

NV HANH Probability and statistics Second semester, 2022-2023 18 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.
6 µ0 .
Case 1.3: We consider the two-tailed test H1 : µ =
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

Find the rejection region Wα : We reject H0 and accept H1 if X̄ is

much different from µ0 or |X̄ − µ0 | is large enough
⇔ |Z | = |X̄σ/−µ |
√ 0 > x, for a critical value x to be found.
n
The rejection region Wα = {|Z | > x} = (−∞; −x) ∪ (x; +∞)
satisfies:
P(Type I error) = P(reject H0 |if H0 is true) = P(Z ∈ Wα |H0 is true)
P(Type I error) = P(|Z | > x|H0 is true) ≤ α ⇔ x = Zα/2
NV HANH Probability and statistics Second semester, 2022-2023 18 / 50
 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

6 µ0 .
Case 1.3: We consider the two-tailed test H1 : µ =
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

The rejection region

Wα = {|Z | > Zα/2 } = (−∞; −Zα/2 ) ∪ (Zα/2 ; +∞).

NV HANH Probability and statistics Second semester, 2022-2023 19 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 1: The population X ∼ N(µ; σ 2 ) where σ 2 is known.

6 µ0 .
Case 1.3: We consider the two-tailed test H1 : µ =
We choose the test statistic
X̄ − µ0
Z= √ ∼ N(0, 1) if H0 is true
σ/ n

The rejection region

Wα = {|Z | > Zα/2 } = (−∞; −Zα/2 ) ∪ (Zα/2 ; +∞).
The decision rule: If |Z | > Zα/2 , then we reject H0 , otherwise we fail
to reject H0 .

NV HANH Probability and statistics Second semester, 2022-2023 19 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 2: The population X ∼ N(µ; σ 2 ) where σ 2 is unknown.

NV HANH Probability and statistics Second semester, 2022-2023 20 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 2: The population X ∼ N(µ; σ 2 ) where σ 2 is unknown.

We choose the test statistic
X̄ − µ0
T = √ ∼ tn−1 if H0 is true
S/ n

NV HANH Probability and statistics Second semester, 2022-2023 20 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 2: The population X ∼ N(µ; σ 2 ) where σ 2 is unknown.

We choose the test statistic
X̄ − µ0
T = √ ∼ tn−1 if H0 is true
S/ n

Similarly, the rejection region is as the following


(tn−1;α ; +∞),
 if H1 : µ > µ0
Wα = (−∞; −tn−1;α ), if H1 : µ < µ0

(−∞; −tn−1;α/2 ) ∪ (tn−1;α/2 ; +∞), if H1 : µ 6= µ0


NV HANH Probability and statistics Second semester, 2022-2023 20 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 3: We do not know X ∼ N(µ; σ 2 ) where σ 2 is known and n is large

enough (n > 30).

NV HANH Probability and statistics Second semester, 2022-2023 21 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 3: We do not know X ∼ N(µ; σ 2 ) where σ 2 is known and n is large

enough (n > 30).
We choose the test statistic
X̄ − µ0
Z= √ ≈ N(0; 1) if H0 is true
σ/ n

NV HANH Probability and statistics Second semester, 2022-2023 21 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 3: We do not know X ∼ N(µ; σ 2 ) where σ 2 is known and n is large

enough (n > 30).
We choose the test statistic
X̄ − µ0
Z= √ ≈ N(0; 1) if H0 is true
σ/ n

Similarly, the rejection region is as the following


(Zα ; +∞),
 if H1 : µ > µ0
Wα = (−∞; −Zα ), if H1 : µ < µ0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : µ 6= µ0


NV HANH Probability and statistics Second semester, 2022-2023 21 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 4: We do not know X ∼ N(µ; σ 2 ) where σ 2 is unknown and n is
large enough (n > 30).

NV HANH Probability and statistics Second semester, 2022-2023 22 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 4: We do not know X ∼ N(µ; σ 2 ) where σ 2 is unknown and n is
large enough (n > 30).
We choose the test statistic
X̄ − µ0
T = √ ≈ N(0; 1) if H0 is true
S/ n

NV HANH Probability and statistics Second semester, 2022-2023 22 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 4: We do not know X ∼ N(µ; σ 2 ) where σ 2 is unknown and n is
large enough (n > 30).
We choose the test statistic
X̄ − µ0
T = √ ≈ N(0; 1) if H0 is true
S/ n

Similarly, the rejection region is as the following


(Zα ; +∞),
 if H1 : µ > µ0
Wα = (−∞; −Zα ), if H1 : µ < µ0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : µ 6= µ0


NV HANH Probability and statistics Second semester, 2022-2023 22 / 50

 Test on a population mean µ.

Test on a population mean µ.

Case 4: We do not know X ∼ N(µ; σ 2 ) where σ 2 is unknown and n is
large enough (n > 30).
We choose the test statistic
X̄ − µ0
T = √ ≈ N(0; 1) if H0 is true
S/ n

Similarly, the rejection region is as the following


(Zα ; +∞),
 if H1 : µ > µ0
Wα = (−∞; −Zα ), if H1 : µ < µ0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : µ 6= µ0


.
Case 5: We do not know X ∼ N(µ; σ 2 ) and n is small. We do not study
this case (the nonparametric test).
NV HANH Probability and statistics Second semester, 2022-2023 22 / 50
 Test on a population mean µ.

Test on a population mean µ.

Example 6.1: Test the hypothesis that the average content of containers
of a particular lubricant differs from 10 liters if the contents of a random
sample of 10 containers are 10.2, 9.7, 10.1, 10.3, 10.1, 9.8, 9.9, 10.4, 10.3,
and 9.8 liters. Use a 0.05 level of significance and assume that the
distribution of contents is normal.

NV HANH Probability and statistics Second semester, 2022-2023 23 / 50

 Test on a population mean µ.

Test on a population mean µ.

Example 6.1: Test the hypothesis that the average content of containers
of a particular lubricant differs from 10 liters if the contents of a random
sample of 10 containers are 10.2, 9.7, 10.1, 10.3, 10.1, 9.8, 9.9, 10.4, 10.3,
and 9.8 liters. Use a 0.05 level of significance and assume that the
distribution of contents is normal.
Let X be the content of containers of lubricant. It is supposed that
X ∼ N(µ; σ 2 ) where σ 2 is unknown. We test the null hypothesis
H0 : µ = 10 against H1 : µ 6= 10 at α = 0.05 level of significance.

NV HANH Probability and statistics Second semester, 2022-2023 23 / 50

 Test on a population mean µ.

Test on a population mean µ.

Example 6.1: Test the hypothesis that the average content of containers
of a particular lubricant differs from 10 liters if the contents of a random
sample of 10 containers are 10.2, 9.7, 10.1, 10.3, 10.1, 9.8, 9.9, 10.4, 10.3,
and 9.8 liters. Use a 0.05 level of significance and assume that the
distribution of contents is normal.
Let X be the content of containers of lubricant. It is supposed that
X ∼ N(µ; σ 2 ) where σ 2 is unknown. We test the null hypothesis
H0 : µ = 10 against H1 : µ 6= 10 at α = 0.05 level of significance.
The test statistic is
X̄ − µ0
T = √ ∼ tn−1 if H0 is true
S/ n

NV HANH Probability and statistics Second semester, 2022-2023 23 / 50

 Test on a population mean µ.

Test on a population mean µ.

For the given sample: n = 10; x̄ = (10.2 + ... + 9.8)/10 = 10.06; s 2 =

(10.22 + ... + 9.82 − 10 ∗ 10.062 )/9 = 0.06; s = 0.245. Then
10.06 − 10
Tobs = √ = 0.774
0.245/ 10

NV HANH Probability and statistics Second semester, 2022-2023 24 / 50

 Test on a population mean µ.

Test on a population mean µ.

For the given sample: n = 10; x̄ = (10.2 + ... + 9.8)/10 = 10.06; s 2 =

(10.22 + ... + 9.82 − 10 ∗ 10.062 )/9 = 0.06; s = 0.245. Then
10.06 − 10
Tobs = √ = 0.774
0.245/ 10

The rejection region is Wα = (−∞; −tn−1;α/2 ) ∪ (tn−1;α/2 ; +∞) =

(−∞; −t9;0.025 ) ∪ (t9;0.025 ; +∞) = (−∞; −2.26) ∪ (2.26; +∞).

NV HANH Probability and statistics Second semester, 2022-2023 24 / 50

 Test on a population mean µ.

Test on a population mean µ.

For the given sample: n = 10; x̄ = (10.2 + ... + 9.8)/10 = 10.06; s 2 =

(10.22 + ... + 9.82 − 10 ∗ 10.062 )/9 = 0.06; s = 0.245. Then
10.06 − 10
Tobs = √ = 0.774
0.245/ 10

The rejection region is Wα = (−∞; −tn−1;α/2 ) ∪ (tn−1;α/2 ; +∞) =

(−∞; −t9;0.025 ) ∪ (t9;0.025 ; +∞) = (−∞; −2.26) ∪ (2.26; +∞).
Since Tobs ∈
/ Wα then we fail to reject H0 .

NV HANH Probability and statistics Second semester, 2022-2023 24 / 50

 Test on a population mean µ.

Test on a population mean µ.

For the given sample: n = 10; x̄ = (10.2 + ... + 9.8)/10 = 10.06; s 2 =

(10.22 + ... + 9.82 − 10 ∗ 10.062 )/9 = 0.06; s = 0.245. Then
10.06 − 10
Tobs = √ = 0.774
0.245/ 10

The rejection region is Wα = (−∞; −tn−1;α/2 ) ∪ (tn−1;α/2 ; +∞) =

(−∞; −t9;0.025 ) ∪ (t9;0.025 ; +∞) = (−∞; −2.26) ∪ (2.26; +∞).
Since Tobs ∈/ Wα then we fail to reject H0 . Thus, there is not enough
evidence to infer that the average content of containers of a lubricant
differs from 10 liters at 5% level of signifiance.

NV HANH Probability and statistics Second semester, 2022-2023 24 / 50

 Test on a population mean µ.

Introduction

Example 6.2: Past experience indicates that the mean of monthly

long-distance telephone bills is µ0 = 17.85 (USD). After an advertising
campaign aimed at increasing long-distance telephone usage, the manager
of the telephone company took a random sample of 25 household bills and
recorded the mean of their monthly usage x̄ = 19.93. It is supposed that
the monthly long-distance telephone bill follows a normal distribution of a
standard deviation σ = 3.39.
Do the data allow us to infer that the campaign was successful at 5%
level of signifiance?
Find the probabillity of type II error of the test if the true mean of
telephone bills after the advertising is 20 (USD).

NV HANH Probability and statistics Second semester, 2022-2023 25 / 50

 Test on a population mean µ.

Test on a population mean µ.

Solution of Example 6.2:

Let X be the monthly long-distance telephone bill of households after
the advertising campaign. It is supposed that X ∼ N(µ; σ 2 ) where
σ = 3.39. We test the null hypothesis H0 : µ = 17.85 against
H1 : µ > 17.85 at α = 0.05 level of significance.

NV HANH Probability and statistics Second semester, 2022-2023 26 / 50

 Test on a population mean µ.

Test on a population mean µ.

Solution of Example 6.2:

Let X be the monthly long-distance telephone bill of households after
the advertising campaign. It is supposed that X ∼ N(µ; σ 2 ) where
σ = 3.39. We test the null hypothesis H0 : µ = 17.85 against
H1 : µ > 17.85 at α = 0.05 level of significance.
The test statistic is
X̄ − µ0
Z= √ ∼ N(0; 1) if H0 is true
σ/ n

NV HANH Probability and statistics Second semester, 2022-2023 26 / 50

 Test on a population mean µ.

Test on a population mean µ.

Solution of Example 6.2:

Let X be the monthly long-distance telephone bill of households after
the advertising campaign. It is supposed that X ∼ N(µ; σ 2 ) where
σ = 3.39. We test the null hypothesis H0 : µ = 17.85 against
H1 : µ > 17.85 at α = 0.05 level of significance.
The test statistic is
X̄ − µ0
Z= √ ∼ N(0; 1) if H0 is true
σ/ n

For the given sample, we have n = 25; x̄ = 19.93 then

19.93 − 17.85
Zobs = √ = 3.07
3.39/ 25

NV HANH Probability and statistics Second semester, 2022-2023 26 / 50

 Test on a population mean µ.

Test on a population mean µ.

The rejection region is

Wα = (Zα ; +∞) = (Z0.05 ; +∞) = (1.645; +∞).

NV HANH Probability and statistics Second semester, 2022-2023 27 / 50

 Test on a population mean µ.

Test on a population mean µ.

The rejection region is

Wα = (Zα ; +∞) = (Z0.05 ; +∞) = (1.645; +∞).
Since Tobs ∈ Wα then we reject H0 . Thus, there is enough evidence
to infer that the campaign was successful at 5% level of signifiance.

NV HANH Probability and statistics Second semester, 2022-2023 27 / 50

 Test on a population mean µ.

Test on a population mean µ.

The rejection region is

Wα = (Zα ; +∞) = (Z0.05 ; +∞) = (1.645; +∞).
Since Tobs ∈ Wα then we reject H0 . Thus, there is enough evidence
to infer that the campaign was successful at 5% level of signifiance.
The probabillity of type II error of the test is

β = P(type II error) = P(accept H0 | H0 is false) = P(Z ∈

/ Wα |µ = 20)
X̄ − µ0
√ ≤ Zα |µ = 20)
= P(Z ≤ Zα |µ = 20) = P(
σ/ n
X̄ − µ µ0 − µ
= P( √ ≤ Zα + √ |µ = 20)
σ/ n σ/ n

NV HANH Probability and statistics Second semester, 2022-2023 27 / 50

 Test on a population mean µ.

Test on a population mean µ.

Since
X̄ − µ
Z= √ ∼ N(0; 1)
σ/ n
for all µ = E (X ),

NV HANH Probability and statistics Second semester, 2022-2023 28 / 50

 Test on a population mean µ.

Test on a population mean µ.

Since
X̄ − µ
Z= √ ∼ N(0; 1)
σ/ n
for all µ = E (X ), then
µ0 − µ 17.85 − 20
β = Φ(Zα + √ ) = Φ(1.645 + √ ) = Φ(−1.53) = 0.063
σ/ n 3.39/ 25

NV HANH Probability and statistics Second semester, 2022-2023 28 / 50

 Test on a population mean µ.

Test on a population mean µ.

Since
X̄ − µ
Z= √ ∼ N(0; 1)
σ/ n
for all µ = E (X ), then
µ0 − µ 17.85 − 20
β = Φ(Zα + √ ) = Φ(1.645 + √ ) = Φ(−1.53) = 0.063
σ/ n 3.39/ 25

The power of test is 1 − β = 1 − 0.063 = 0.937.

NV HANH Probability and statistics Second semester, 2022-2023 28 / 50

 Test on a population proportion p.

Test on a population proportion p.

Example: Suppose that the proportion of defective items in a production

line is 1.2%. A technical improvement was made to reduce the rate of
defective items. Test 1000 items from the production line after the
improvement, 8 of them were found to be defective. Do the data allow us
to infer that the improvement was successful at 10% level of signifiance?

NV HANH Probability and statistics Second semester, 2022-2023 29 / 50

 Test on a population proportion p.

Test on a population proportion p.

Example: Suppose that the proportion of defective items in a production

line is 1.2%. A technical improvement was made to reduce the rate of
defective items. Test 1000 items from the production line after the
improvement, 8 of them were found to be defective. Do the data allow us
to infer that the improvement was successful at 10% level of signifiance?
Let p be the proportion of defective items in the production line after
the technical improvement.

NV HANH Probability and statistics Second semester, 2022-2023 29 / 50

 Test on a population proportion p.

Test on a population proportion p.

Example: Suppose that the proportion of defective items in a production

line is 1.2%. A technical improvement was made to reduce the rate of
defective items. Test 1000 items from the production line after the
improvement, 8 of them were found to be defective. Do the data allow us
to infer that the improvement was successful at 10% level of signifiance?
Let p be the proportion of defective items in the production line after
the technical improvement.
We test the research hypthesis that p < 0.012

NV HANH Probability and statistics Second semester, 2022-2023 29 / 50

 Test on a population proportion p.

Test on a population proportion p.

Problem 2: Let p be a poppulation proportion (for example, the proportion
of defective items in a production line).

NV HANH Probability and statistics Second semester, 2022-2023 30 / 50

 Test on a population proportion p.

Test on a population proportion p.

Problem 2: Let p be a poppulation proportion (for example, the proportion
of defective items in a production line).
We test the null hypotheses H0 : p = p0 against the alternative
H1 : p > p0 (or <, 6=) at level of significance α.

NV HANH Probability and statistics Second semester, 2022-2023 30 / 50

 Test on a population proportion p.

Test on a population proportion p.

Problem 2: Let p be a poppulation proportion (for example, the proportion
of defective items in a production line).
We test the null hypotheses H0 : p = p0 against the alternative
H1 : p > p0 (or <, 6=) at level of significance α.
Take a random sample of size n (large enough) from the population.
Let p̂ be the sample proportion (point estimator of p). The test
statistic is
p̂ − p0
Z=p ∼ N(0; 1) if H0 is true
p0 (1 − p0 )/n

NV HANH Probability and statistics Second semester, 2022-2023 30 / 50

 Test on a population proportion p.

Test on a population proportion p.

Problem 2: Let p be a poppulation proportion (for example, the proportion
of defective items in a production line).
We test the null hypotheses H0 : p = p0 against the alternative
H1 : p > p0 (or <, 6=) at level of significance α.
Take a random sample of size n (large enough) from the population.
Let p̂ be the sample proportion (point estimator of p). The test
statistic is
p̂ − p0
Z=p ∼ N(0; 1) if H0 is true
p0 (1 − p0 )/n

The rejection region is as the following


(Zα ; +∞),
 if H1 : p > p0
Wα = (−∞; −Zα ), if H1 : p < p0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : p 6= p0

.
NV HANH Probability and statistics Second semester, 2022-2023 30 / 50
 Test on a population proportion p.

Test on a population proportion p.

Example 6.3: Suppose that the proportion of defective items in a

production line is 1.2%. A technical improvement was made to reduce the
rate of defective items. Test 1000 items from the production line after the
improvement, 8 of them were found to be defective.
Do the data allow us to infer that the improvement was successful at
10% level of signifiance?
Find the power of test if the true proportion of defective items after
the technical improvement is p1 = 0.5%.

NV HANH Probability and statistics Second semester, 2022-2023 31 / 50

 Test on a population proportion p.

Test on a population proportion p.

Solution:
Let p be the proportion of defective items in the production line after
the technical improvement. We test the null hypotheses
H0 : p = 0.012 against the alternative H1 : p < 0.012 at level of
significance α = 0.1.

NV HANH Probability and statistics Second semester, 2022-2023 32 / 50

 Test on a population proportion p.

Test on a population proportion p.

Solution:
Let p be the proportion of defective items in the production line after
the technical improvement. We test the null hypotheses
H0 : p = 0.012 against the alternative H1 : p < 0.012 at level of
significance α = 0.1.
The test statistic is
p̂ − p0
Z=p ∼ N(0; 1) if H0 is true
p0 (1 − p0 )/n

NV HANH Probability and statistics Second semester, 2022-2023 32 / 50

 Test on a population proportion p.

Test on a population proportion p.

Solution:
Let p be the proportion of defective items in the production line after
the technical improvement. We test the null hypotheses
H0 : p = 0.012 against the alternative H1 : p < 0.012 at level of
significance α = 0.1.
The test statistic is
p̂ − p0
Z=p ∼ N(0; 1) if H0 is true
p0 (1 − p0 )/n
For the given sample: n = 1000; p̂ = 8/1000 = 0.008, p0 = 0.012 then
0.008 − 0.012
Zobs = p = −1.42
0.008 ∗ 0.992/1000

NV HANH Probability and statistics Second semester, 2022-2023 32 / 50

 Test on a population proportion p.

Test on a population proportion p.

Solution:
Let p be the proportion of defective items in the production line after
the technical improvement. We test the null hypotheses
H0 : p = 0.012 against the alternative H1 : p < 0.012 at level of
significance α = 0.1.
The test statistic is
p̂ − p0
Z=p ∼ N(0; 1) if H0 is true
p0 (1 − p0 )/n
For the given sample: n = 1000; p̂ = 8/1000 = 0.008, p0 = 0.012 then
0.008 − 0.012
Zobs = p = −1.42
0.008 ∗ 0.992/1000

The rejection region is Wα = (−∞; −Zα ) = (−∞; −1.28).

NV HANH Probability and statistics Second semester, 2022-2023 32 / 50

 Test on a population proportion p.

Test on a population proportion p.

Solution:
Let p be the proportion of defective items in the production line after
the technical improvement. We test the null hypotheses
H0 : p = 0.012 against the alternative H1 : p < 0.012 at level of
significance α = 0.1.
The test statistic is
p̂ − p0
Z=p ∼ N(0; 1) if H0 is true
p0 (1 − p0 )/n
For the given sample: n = 1000; p̂ = 8/1000 = 0.008, p0 = 0.012 then
0.008 − 0.012
Zobs = p = −1.42
0.008 ∗ 0.992/1000

The rejection region is Wα = (−∞; −Zα ) = (−∞; −1.28).

Since Zobs ∈ Wα then we reject H0 . Thus, there is enough evidence
to infer that the technical improvement was successful at 10% level of
signifiance.
NV HANH Probability and statistics Second semester, 2022-2023 32 / 50
 Test on a population proportion p.

Test on a population proportion p.

The power of the test is

1 − β = P(reject H0 | H0 is false) = P(Z ∈ Wα |p = p1 )

p̂ − p0
= P(Z < −Zα |p = p1 ) = P( p < −Zα |p = p1 )
p0 (1 − p0 )/n

NV HANH Probability and statistics Second semester, 2022-2023 33 / 50

 Test on a population proportion p.

Test on a population proportion p.

The power of the test is

1 − β = P(reject H0 | H0 is false) = P(Z ∈ Wα |p = p1 )

p̂ − p0
= P(Z < −Zα |p = p1 ) = P( p < −Zα |p = p1 )
p0 (1 − p0 )/n
s
p̂ − p1 p0 (1 − p0 ) p0 − p1
= P( p < −Zα +p |p = p1 )
p1 (1 − p1 )/n p1 (1 − p1 ) p1 (1 − p1 )/n

NV HANH Probability and statistics Second semester, 2022-2023 33 / 50

 Test on a population proportion p.

Test on a population proportion p.

The power of the test is

1 − β = P(reject H0 | H0 is false) = P(Z ∈ Wα |p = p1 )

p̂ − p0
= P(Z < −Zα |p = p1 ) = P( p < −Zα |p = p1 )
p0 (1 − p0 )/n
s
p̂ − p1 p0 (1 − p0 ) p0 − p1
= P( p < −Zα +p |p = p1 )
p1 (1 − p1 )/n p1 (1 − p1 ) p1 (1 − p1 )/n

p̂−p1
Since Z = √ ∼ N(0, 1) if p = p1 , thus
p1 (1−p1 )/n
s
p0 (1 − p0 ) p0 − p1
1 − β = Φ(−Zα +p )
p1 (1 − p1 ) p1 (1 − p1 )/n

NV HANH Probability and statistics Second semester, 2022-2023 33 / 50

 Test on a population proportion p.

Test on a population proportion p.

When p1 = 0.005, n = 1000, p0 = 0.012, Zα = 1.28, the power of the

test is
r
0.012 ∗ 0.988 0.012 − 0.005
1 − β = Φ(−1.28 +p )
0.005 ∗ 0.995 0.005 ∗ 0.995/1000
= Φ(1.16) = 0.88

NV HANH Probability and statistics Second semester, 2022-2023 34 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Example: A high school math teacher claims that students in her class will
score higher on the math portion of the ACT than students in a colleague’s
math class. The mean ACT math score for 49 students in her class is 22.1
and the sample standard deviation is 4.8. The mean ACT math score for
44 of the colleague’s students is 19.8 and the sample standard deviation is
5.4. At α = 0.1, can the teacher’s claim be supported?

NV HANH Probability and statistics Second semester, 2022-2023 35 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Example: A high school math teacher claims that students in her class will
score higher on the math portion of the ACT than students in a colleague’s
math class. The mean ACT math score for 49 students in her class is 22.1
and the sample standard deviation is 4.8. The mean ACT math score for
44 of the colleague’s students is 19.8 and the sample standard deviation is
5.4. At α = 0.1, can the teacher’s claim be supported?
Let µ1 , µ2 be the mean ACT math scores of students in her class and
in her colleague’s class, respectively.
We want to test whether µ1 − µ2 > 0 at α = 0.1 level of significance.

NV HANH Probability and statistics Second semester, 2022-2023 35 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Problem 3: We observe 2 poppulations X1 , X2 where µ1 = E (X1 ),

µ2 = E (X2 ) and σ12 = V (X1 ), σ22 = V (X2 ).

NV HANH Probability and statistics Second semester, 2022-2023 36 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Problem 3: We observe 2 poppulations X1 , X2 where µ1 = E (X1 ),

µ2 = E (X2 ) and σ12 = V (X1 ), σ22 = V (X2 ).
We test on the difference between two means µ1 − µ2 .

NV HANH Probability and statistics Second semester, 2022-2023 36 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Problem 3: We observe 2 poppulations X1 , X2 where µ1 = E (X1 ),

µ2 = E (X2 ) and σ12 = V (X1 ), σ22 = V (X2 ).
We test on the difference between two means µ1 − µ2 .
The null hypothesis is H0 : µ1 − µ2 = 0 against the alternative
H1 : µ1 − µ2 > 0 (or >, 6=).

NV HANH Probability and statistics Second semester, 2022-2023 36 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Problem 3: We observe 2 poppulations X1 , X2 where µ1 = E (X1 ),

µ2 = E (X2 ) and σ12 = V (X1 ), σ22 = V (X2 ).
We test on the difference between two means µ1 − µ2 .
The null hypothesis is H0 : µ1 − µ2 = 0 against the alternative
H1 : µ1 − µ2 > 0 (or >, 6=).
Consider 2 independent random sample of size n1 , n2 from 2
populations. Let X̄1 , X̄2 be the sample means and S12 , S22 be the
sample variances of 2 samples, respectively. The sampling distribution
of X̄1 − X̄2 is the following:

X̄1 − X̄2 − (µ1 − µ2 )

Z= q 2 ∼ N(0, 1)
σ1 σ22
n1 + n2

NV HANH Probability and statistics Second semester, 2022-2023 36 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 1: Suppose that 2 populations X1 , X2 are normal and 2 variances

σ12 , σ22 are known.

NV HANH Probability and statistics Second semester, 2022-2023 37 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 1: Suppose that 2 populations X1 , X2 are normal and 2 variances

σ12 , σ22 are known.
The test statistic is
X̄1 − X̄2
Z=q 2 ∼ N(0, 1) if H0 is true
σ1 σ22
n1 + n2

NV HANH Probability and statistics Second semester, 2022-2023 37 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 1: Suppose that 2 populations X1 , X2 are normal and 2 variances

σ12 , σ22 are known.
The test statistic is
X̄1 − X̄2
Z=q 2 ∼ N(0, 1) if H0 is true
σ1 σ22
n1 + n2

The rejection region is as the following


(Zα ; +∞),
 if H1 : µ1 − µ2 > 0
Wα = (−∞; −Zα ), if H1 : µ1 − µ2 < 0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : µ1 − µ2 6= 0


NV HANH Probability and statistics Second semester, 2022-2023 37 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 2: Two populations X1 , X2 are not supposed to be normal, 2 variances

σ12 , σ22 are unknown and the sample sizes n1 , n2 are large enough (> 30).

NV HANH Probability and statistics Second semester, 2022-2023 38 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 2: Two populations X1 , X2 are not supposed to be normal, 2 variances

σ12 , σ22 are unknown and the sample sizes n1 , n2 are large enough (> 30).
The test statistic is
X̄1 − X̄2
T =q 2 ≈ N(0, 1) if H0 is true
S1 S22
n1 + n2

NV HANH Probability and statistics Second semester, 2022-2023 38 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 2: Two populations X1 , X2 are not supposed to be normal, 2 variances

σ12 , σ22 are unknown and the sample sizes n1 , n2 are large enough (> 30).
The test statistic is
X̄1 − X̄2
T =q 2 ≈ N(0, 1) if H0 is true
S1 S22
n1 + n2

The rejection region is as the following


(Zα ; +∞),
 if H1 : µ1 − µ2 > 0
Wα = (−∞; −Zα ), if H1 : µ1 − µ2 < 0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : µ1 − µ2 6= 0


NV HANH Probability and statistics Second semester, 2022-2023 38 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 3: Two populations X1 , X2 are supposed to be normal, 2 variances
σ12 , σ22 are unknown and equal (σ12 = σ22 ).

NV HANH Probability and statistics Second semester, 2022-2023 39 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 3: Two populations X1 , X2 are supposed to be normal, 2 variances
σ12 , σ22 are unknown and equal (σ12 = σ22 ).
The test statistic is
X̄1 − X̄2
T =q ∼ td.f if H0 is true,
S2 S2
n1 + n2

where d.f = n1 + n2 − 2 and

(n1 − 1)S12 + (n2 − 1)S22
S2 = is called the pooled variance.
n1 + n2 − 2

NV HANH Probability and statistics Second semester, 2022-2023 39 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2

Case 3: Two populations X1 , X2 are supposed to be normal, 2 variances
σ12 , σ22 are unknown and equal (σ12 = σ22 ).
The test statistic is
X̄1 − X̄2
T =q ∼ td.f if H0 is true,
S2 S2
n1 + n2

where d.f = n1 + n2 − 2 and

(n1 − 1)S12 + (n2 − 1)S22
S2 = is called the pooled variance.
n1 + n2 − 2

The rejection region is as the following


(td.f ,α ; +∞),
 if H1 : µ1 − µ2 > 0
Wα = (−∞; −td.f ,α ), if H1 : µ1 − µ2 < 0

(−∞; −td.f ,α/2 ) ∪ (td.f ,α/2 ; +∞), if H1 : µ1 − µ2 6= 0.

.
NV HANH Probability and statistics Second semester, 2022-2023 39 / 50
 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Example 6.4: A high school math teacher claims that students in her class
will score higher on the math portion of the ACT than students in a
colleague’s math class. The mean ACT math score for 49 students in her
class is 22.1 and the sample standard deviation is 4.8. The mean ACT
math score for 44 of the colleague’s students is 19.8 and the sample
standard deviation is 5.4. At α = 0.1, can the teacher’s claim be
supported?

NV HANH Probability and statistics Second semester, 2022-2023 40 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Example 6.4: A high school math teacher claims that students in her class
will score higher on the math portion of the ACT than students in a
colleague’s math class. The mean ACT math score for 49 students in her
class is 22.1 and the sample standard deviation is 4.8. The mean ACT
math score for 44 of the colleague’s students is 19.8 and the sample
standard deviation is 5.4. At α = 0.1, can the teacher’s claim be
supported?
Solution:
Let µ1 , µ2 be the mean ACT math scores of students in her class and
in her colleague’s class, respectively. We test H0 : µ1 − µ2 = 0 versus
H1 : µ1 − µ2 > 0 at α = 0.1 level of significance.

NV HANH Probability and statistics Second semester, 2022-2023 40 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

It is not supposed that the populations are normal and the two
population variances are unknown, but the sample sizes
n1 = 49, n2 = 44 are large enough. Thus, the test statistic is

X̄1 − X̄2
T =q 2 ≈ N(0, 1) if H0 is true
S1 S22
n1 + n2

NV HANH Probability and statistics Second semester, 2022-2023 41 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

It is not supposed that the populations are normal and the two
population variances are unknown, but the sample sizes
n1 = 49, n2 = 44 are large enough. Thus, the test statistic is

X̄1 − X̄2
T =q 2 ≈ N(0, 1) if H0 is true
S1 S22
n1 + n2

For the given sample, we have: n1 = 49, n2 = 44; x̄1 = 22.1,

x̄2 = 19.8, s1 = 4.8, s2 = 5.4, then
22.1 − 19.8
Tobs = q = 2.16
4.82 5.42
49 + 44

NV HANH Probability and statistics Second semester, 2022-2023 41 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

The rejection region is Wα = (Zα ; +∞) = (Z0.1 ; +∞) = (1.28; +∞).

NV HANH Probability and statistics Second semester, 2022-2023 42 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

The rejection region is Wα = (Zα ; +∞) = (Z0.1 ; +∞) = (1.28; +∞).

Since Tobs = 2.16 ∈ Wα then we reject H0 . Thus, there is enough
evidence at the 10% level to support the teacher’s claim that her
students score better on the ACT.

NV HANH Probability and statistics Second semester, 2022-2023 42 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Example 6.5: To find out whether a new serum will arrest leukemia, 9
mice, all with an advanced stage of the disease, are selected. Five mice
receive the treatment and 4 do not. Survival times, in years, from the time
the experiment commenced are as follows:
Treatment: 2.1, 5.3, 1.4, 4.6, 0.9
No Treatment: 1.9, 0.5, 2.8, 3.1
At the 0.05 level of significance, can the serum be said to be effective?
Assume the two populations to be normally distributed with equal
variances.

NV HANH Probability and statistics Second semester, 2022-2023 43 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Example 6.5: To find out whether a new serum will arrest leukemia, 9
mice, all with an advanced stage of the disease, are selected. Five mice
receive the treatment and 4 do not. Survival times, in years, from the time
the experiment commenced are as follows:
Treatment: 2.1, 5.3, 1.4, 4.6, 0.9
No Treatment: 1.9, 0.5, 2.8, 3.1
At the 0.05 level of significance, can the serum be said to be effective?
Assume the two populations to be normally distributed with equal
variances.
Solution:
Let X1 , X2 be the survival times, in years, of mice with and without
treatment of the new serum. To test whether the serum is said to be
effective, we test H0 : µ1 − µ2 = 0 versus H1 : µ1 − µ2 > 0 at
α = 0.05 level of significance.

NV HANH Probability and statistics Second semester, 2022-2023 43 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

It is supposed that X1 ∼ N(µ1 , σ12 ), X2 ∼ N(µ2 , σ22 ) where σ1 , σ2 are

unknown but equal. Thus, the test statistic is

X̄1 − X̄2
T =q ∼ tn1 +n2 −2 if H0 is true
S2 S2
n1 + n2

NV HANH Probability and statistics Second semester, 2022-2023 44 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

It is supposed that X1 ∼ N(µ1 , σ12 ), X2 ∼ N(µ2 , σ22 ) where σ1 , σ2 are

unknown but equal. Thus, the test statistic is

X̄1 − X̄2
T =q ∼ tn1 +n2 −2 if H0 is true
S2 S2
n1 + n2

For the given sample, we have: n1 = 5, n2 = 4,

x̄1 = (2.1 + 5.3 + 1.4 + 4.6 + 0.9)/5 = 2.86,
x̄2 = (1.9 + 0.5 + 2.8 + 3.1)/4 = 2.075,
s12 = (2.12 + .. + 0.92 − 5 ∗ 2.862 )/4 = 3.883,
s22 = (1.92 + ... + 3.12 − 4 ∗ 2.0752 )/3 = 1.3625, then

(n1 − 1)s12 + (n2 − 1)s22 4 ∗ 3.883 + 3 ∗ 1.3625

s2 = = = 2.803
n1 + n2 − 2 7

NV HANH Probability and statistics Second semester, 2022-2023 44 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Thus
2.86 − 2.075
Tobs = q = 0.699
2.803 2.803
5 + 4

NV HANH Probability and statistics Second semester, 2022-2023 45 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Thus
2.86 − 2.075
Tobs = q = 0.699
2.803 2.803
5 + 4

The rejection region is

Wα = (tn1 +n2 −2;α ; +∞) = (t7;0.05 ; +∞) = (1.89; +∞).

NV HANH Probability and statistics Second semester, 2022-2023 45 / 50

 Test on the difference between two means µ1 − µ2 .

Test on the difference between two means µ1 − µ2 .

Thus
2.86 − 2.075
Tobs = q = 0.699
2.803 2.803
5 + 4

The rejection region is

Wα = (tn1 +n2 −2;α ; +∞) = (t7;0.05 ; +∞) = (1.89; +∞).
Since Tobs = 0.699 ∈
/ Wα then we fail to reject H0 . Thus, there is not
enough evidence at the 5% level to infer that the new serum is said to
be effective.

NV HANH Probability and statistics Second semester, 2022-2023 45 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Problem 4: Let p1 , p2 be two population proportions (for example,
proportions of defecctive items in 2 different production lines).

NV HANH Probability and statistics Second semester, 2022-2023 46 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Problem 4: Let p1 , p2 be two population proportions (for example,
proportions of defecctive items in 2 different production lines). We
consider the following test: H0 : p1 − p2 = 0 against H1 : p1 − p2 > 0 (or
>, 6=).

NV HANH Probability and statistics Second semester, 2022-2023 46 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Problem 4: Let p1 , p2 be two population proportions (for example,
proportions of defecctive items in 2 different production lines). We
consider the following test: H0 : p1 − p2 = 0 against H1 : p1 − p2 > 0 (or
>, 6=).
Draw 2 independent random samples of size n1 , n2 from 2 populations.

NV HANH Probability and statistics Second semester, 2022-2023 46 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Problem 4: Let p1 , p2 be two population proportions (for example,
proportions of defecctive items in 2 different production lines). We
consider the following test: H0 : p1 − p2 = 0 against H1 : p1 − p2 > 0 (or
>, 6=).
Draw 2 independent random samples of size n1 , n2 from 2 populations.
Let p̂1 , p̂2 be two sample proportions, the point estimates of p1 , p2 ,
respectively.

NV HANH Probability and statistics Second semester, 2022-2023 46 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Problem 4: Let p1 , p2 be two population proportions (for example,
proportions of defecctive items in 2 different production lines). We
consider the following test: H0 : p1 − p2 = 0 against H1 : p1 − p2 > 0 (or
>, 6=).
Draw 2 independent random samples of size n1 , n2 from 2 populations.
Let p̂1 , p̂2 be two sample proportions, the point estimates of p1 , p2 ,
respectively.
The test statistic is
p̂1 − p̂2
Z=r   ≈ N(0, 1) if H0 is true,
1 1
p̂(1 − p̂) n1 + n2

where p̂ is the weighted estimate of p1 and p2 :

n1 p̂1 + n2 p̂2
p̂ =
n1 + n2
NV HANH Probability and statistics Second semester, 2022-2023 46 / 50
 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2

Remark that
m1 m2 m1 + m2
p̂1 = ; p̂2 = ; p̂ = ,
n1 n2 n1 + n2
where m1 , m2 are the number of successes in 2 samples, respectively.

NV HANH Probability and statistics Second semester, 2022-2023 47 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2

Remark that
m1 m2 m1 + m2
p̂1 = ; p̂2 = ; p̂ = ,
n1 n2 n1 + n2
where m1 , m2 are the number of successes in 2 samples, respectively.
The rejection region is as the following

(Zα ; +∞),
 if H1 : p1 − p2 > 0
Wα = (−∞; −Zα ), if H1 : p1 − p2 < 0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : p1 − p2 6= 0


NV HANH Probability and statistics Second semester, 2022-2023 47 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2

Remark that
m1 m2 m1 + m2
p̂1 = ; p̂2 = ; p̂ = ,
n1 n2 n1 + n2
where m1 , m2 are the number of successes in 2 samples, respectively.
The rejection region is as the following

(Zα ; +∞),
 if H1 : p1 − p2 > 0
Wα = (−∞; −Zα ), if H1 : p1 − p2 < 0

(−∞; −Zα/2 ) ∪ (Zα/2 ; +∞), if H1 : p1 − p2 6= 0


.
The sample sizes n1 , n2 must be large enough (we usually suppose
that n1 p1 , n1 (1 − p1 ), n2 p2 , n2 (1 − p2 ) are ≥ 5).

NV HANH Probability and statistics Second semester, 2022-2023 47 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Example 6.5: A recent survey stated that male college students smoke less
than female college students. In a survey of 1245 male students, 361 said
they smoke at least one pack of cigarettes a day. In a survey of 1065
female students, 341 said they smoke at least one pack a day. At α = 0.1,
can you support the claim that the proportion of male college students
who smoke at least one pack of cigarettes a day is lower than the
proportion of female college students who smoke at least one pack a day?

NV HANH Probability and statistics Second semester, 2022-2023 48 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

Example 6.5: A recent survey stated that male college students smoke less
than female college students. In a survey of 1245 male students, 361 said
they smoke at least one pack of cigarettes a day. In a survey of 1065
female students, 341 said they smoke at least one pack a day. At α = 0.1,
can you support the claim that the proportion of male college students
who smoke at least one pack of cigarettes a day is lower than the
proportion of female college students who smoke at least one pack a day?
Solution:
Let p1 , p2 be the proportions of male and of female college students
who smoke at least one pack of cigarettes a day, respectively. We test
H0 : p1 − p2 = 0 versus H1 : p1 − p2 < 0 at α = 0.1 level of
significance.

NV HANH Probability and statistics Second semester, 2022-2023 48 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

The test statistic is

p̂1 − p̂2
Z=r   ≈ N(0, 1) if H0 is true.
p̂(1 − p̂) n11 + 1
n2

NV HANH Probability and statistics Second semester, 2022-2023 49 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

The test statistic is

p̂1 − p̂2
Z=r   ≈ N(0, 1) if H0 is true.
p̂(1 − p̂) n11 + 1
n2

For the given samples, we have: n1 = 1245, n2 = 1065;

m1 = 361, m2 = 341,

NV HANH Probability and statistics Second semester, 2022-2023 49 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

The test statistic is

p̂1 − p̂2
Z=r   ≈ N(0, 1) if H0 is true.
p̂(1 − p̂) n11 + 1
n2

For the given samples, we have: n1 = 1245, n2 = 1065;

m1 = 361, m2 = 341, then
361 341 361 + 341
p̂1 = = 0.29; p̂2 = = 0.32; p̂ = = 0.304
1245 1065 1245 + 1065

NV HANH Probability and statistics Second semester, 2022-2023 49 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

The test statistic is

p̂1 − p̂2
Z=r   ≈ N(0, 1) if H0 is true.
p̂(1 − p̂) n11 + 1
n2

For the given samples, we have: n1 = 1245, n2 = 1065;

m1 = 361, m2 = 341, then
361 341 361 + 341
p̂1 = = 0.29; p̂2 = = 0.32; p̂ = = 0.304
1245 1065 1245 + 1065
Thus,
0.29 − 0.32
Z=r   = −1.56
1 1
0.304 ∗ 0.694 1245 + 1065

NV HANH Probability and statistics Second semester, 2022-2023 49 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

The rejection region is

Wα = (−∞; −Zα ) = (−∞; Z0.1 ) = (−∞; −1.28).

NV HANH Probability and statistics Second semester, 2022-2023 50 / 50

 Test on the difference between two proportions p1 − p2

Test on the difference between two proportions p1 − p2 .

The rejection region is

Wα = (−∞; −Zα ) = (−∞; Z0.1 ) = (−∞; −1.28).
Since Zobs = −1.56 ∈ Wα then we reject H0 . Thus, there is enough
evidence at the 10% level to claim that the proportion of male college
students who smoke at least one pack of cigarettes a day is lower
than the proportion of female college students who smoke at least one
pack a day.

NV HANH Probability and statistics Second semester, 2022-2023 50 / 50