STATISTICAL INFERENCE FOR TWO SAMPLES

ELECTRONIC VERSION OF LECTURE

Dr. Lê Xuân Đại

HoChiMinh City University of Technology
Faculty of Applied Science, Department of Applied Mathematics
Email: ytkadai@[Link]

HCMC — 2025.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 1 / 33
CONTENT

1 HYPOTHESIS TESTING FOR THE MEANS OF TWO INDEPENDENT

SAMPLES

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 2 / 33
CONTENT

1 HYPOTHESIS TESTING FOR THE MEANS OF TWO INDEPENDENT

SAMPLES

2 COMPARISON OF TWO MEANS WITH UNKNOWN VARIANCES

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 2 / 33
CONTENT

1 HYPOTHESIS TESTING FOR THE MEANS OF TWO INDEPENDENT

SAMPLES

2 COMPARISON OF TWO MEANS WITH UNKNOWN VARIANCES

3 COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION WITH
L ARGE SAMPLE SIZES

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 2 / 33
CONTENT

1 HYPOTHESIS TESTING FOR THE MEANS OF TWO INDEPENDENT

SAMPLES

2 COMPARISON OF TWO MEANS WITH UNKNOWN VARIANCES

3 COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION WITH
L ARGE SAMPLE SIZES

4 COMPARISON OF TWO POPULATION PROPORTIONS

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 2 / 33
CONTENT

1 HYPOTHESIS TESTING FOR THE MEANS OF TWO INDEPENDENT

SAMPLES

2 COMPARISON OF TWO MEANS WITH UNKNOWN VARIANCES

3 COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION WITH
L ARGE SAMPLE SIZES

4 COMPARISON OF TWO POPULATION PROPORTIONS

5 PAIRED SAMPLE TESTING

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 2 / 33
PROBLEM

The metal content in drinking water affects public

health. A study analyzed 6 water samples from
different rivers to determine zinc concentrations
(mg/L) in surface water and bottom water. The
recorded data is as follows:
Zinc 1 2 3 4 5 6
Bottom 0.430 0.266 0.567 0.531 0.707 0.716
Surface 0.415 0.238 0.390 0.410 0.605 0.609
Question: Is there a significant difference between
the average zinc concentration in bottom water and
surface water?
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 3 / 33
DEFINITION 0.1
1
Two samples are independent if the
sample chosen from the first population
has no relation to the sample chosen from
the second population.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 4 / 33
DEFINITION 0.1
1
Two samples are independent if the
sample chosen from the first population
has no relation to the sample chosen from
the second population.
2
Two samples are dependent if each
observation in one sample is correlated
with an observation in the other sample.
Dependent samples are often referred to
as paired samples.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 4 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 which follows a normal distribution
N (µ1 , σ21 ) with known variance σ21

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 5 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 which follows a normal distribution
N (µ1 , σ21 ) with known variance σ21
2
Y1 , . . . , Ym are random samples selected from
population 2 which follows a normal distribution
N (µ2 , σ22 ) with known variance σ22

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 5 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 which follows a normal distribution
N (µ1 , σ21 ) with known variance σ21
2
Y1 , . . . , Ym are random samples selected from
population 2 which follows a normal distribution
N (µ2 , σ22 ) with known variance σ22
3
Population 1 and population 2 (represented by X
and Y ) are independent of each other.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 5 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 which follows a normal distribution
N (µ1 , σ21 ) with known variance σ21
2
Y1 , . . . , Ym are random samples selected from
population 2 which follows a normal distribution
N (µ2 , σ22 ) with known variance σ22
3
Population 1 and population 2 (represented by X
and Y ) are independent of each other.
4
E (X − Y ) = E (X ) − E (Y ) = µ1 − µ2 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 5 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 which follows a normal distribution
N (µ1 , σ21 ) with known variance σ21
2
Y1 , . . . , Ym are random samples selected from
population 2 which follows a normal distribution
N (µ2 , σ22 ) with known variance σ22
3
Population 1 and population 2 (represented by X
and Y ) are independent of each other.
4
E (X − Y ) = E (X ) − E (Y ) = µ1 − µ2 .
σ21 σ22
5
V (X − Y ) = V (X ) + V (Y ) = +
n m
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 5 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 6 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2
x − y − (µ1 − µ2 )
Compute the test statistic: z 0 = s
σ21 σ22
+
n m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 6 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2
x − y − (µ1 − µ2 )
Compute the test statistic: z 0 = s
σ21 σ22
+
n m
For a given significance level α, determine the
rejection region and P -value according to the
rule:
Alternative Rejection Region P -value
µ1 ̸= µ2 Wα = {z 0 : |z 0 | > z α/2 } 2[1 − Φ(|z 0 |)]
µ1 > µ2 Wα = {z 0 : z 0 > z α } 1 − Φ(z 0 )
µ1 < µ2 Wα = {z 0 : z 0 < −z α } Φ(z 0 )

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 6 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

EXAMPLE 1.1
A random sample of n = 20 cold-rolled steel specimens
is analyzed to determine hardness, yielding a sample
mean of x = 29.8. A second random sample of m = 25
galvanized steel specimens (coated on both sides)
gives a sample mean hardness of y = 34.7. Assume the
measurements are normally distributed with
standard deviations σ1 = 4 and σ2 = 5, respectively. Is
there sufficient evidence to conclude that µ1 ̸= µ2 at
the significance level α = 0.05?

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 7 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

HYPOTHESIS TESTING STEPS

1 H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 8 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

HYPOTHESIS TESTING STEPS

1 H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
x − y − (µ1 − µ2 ) 29.8 − 34.7 − 0
2 z0 = s = s ≈ −3.6522
σ1 σ2
2 2
4 2
5 2
+ +
n m 20 25

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 8 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

HYPOTHESIS TESTING STEPS

1 H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
x − y − (µ1 − µ2 ) 29.8 − 34.7 − 0
2 z0 = s = s ≈ −3.6522
σ1 σ2
2 2
4 2
5 2
+ +
n m 20 25
3 With α = 0.05 ⇒ z α/2 = z 0.025 = 1.96. Therefore, since
|z 0 | > z α/2 , we reject H0 , which means there is sufficient
evidence to conclude that µ1 is different from µ2 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 8 / 33
 Comparing the means of two normal distributions with known
Hypothesis testing for the means of two independent samples variances

HYPOTHESIS TESTING STEPS

1 H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
x − y − (µ1 − µ2 ) 29.8 − 34.7 − 0
2 z0 = s = s ≈ −3.6522
σ1 σ2
2 2
4 2
5 2
+ +
n m 20 25
3 With α = 0.05 ⇒ z α/2 = z 0.025 = 1.96. Therefore, since
|z 0 | > z α/2 , we reject H0 , which means there is sufficient
evidence to conclude that µ1 is different from µ2 .
4 P = 2[1 − Φ(|z 0 |)] = 2[1 − Φ(3.6522)] ≈ 0.00026. Hence, since
P < 0.05 = α, we reject H0 , which means there is sufficient
evidence to conclude that µ1 is different from µ2 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 8 / 33
 Comparison of Two Means with Unknown Variances

Assumptions:
1 X 1 , . . . , X n are random samples selected from population 1
which follows a normal distribution N (µ1 , σ21 ) with
unknown variance σ21 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 9 / 33
 Comparison of Two Means with Unknown Variances

Assumptions:
1 X 1 , . . . , X n are random samples selected from population 1
which follows a normal distribution N (µ1 , σ21 ) with
unknown variance σ21 .
2 Y1 , . . . , Ym are random samples selected from population 2
which follows a normal distribution N (µ2 , σ22 ) with
unknown variance σ22 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 9 / 33
 Comparison of Two Means with Unknown Variances

Assumptions:
1 X 1 , . . . , X n are random samples selected from population 1
which follows a normal distribution N (µ1 , σ21 ) with
unknown variance σ21 .
2 Y1 , . . . , Ym are random samples selected from population 2
which follows a normal distribution N (µ2 , σ22 ) with
unknown variance σ22 .
3 Populations 1 and 2 are independent of each other.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 9 / 33
 Comparison of Two Means with Unknown Variances

Assumptions:
1 X 1 , . . . , X n are random samples selected from population 1
which follows a normal distribution N (µ1 , σ21 ) with
unknown variance σ21 .
2 Y1 , . . . , Ym are random samples selected from population 2
which follows a normal distribution N (µ2 , σ22 ) with
unknown variance σ22 .
3 Populations 1 and 2 are independent of each other.
Using the null hypothesis and the sample standard deviations
(unbiased) from population 1 and 2, denoted by s 1 and s 2 :
s1
1 If ∈ [0.5; 2], then assume σ21 = σ22
s2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 9 / 33
 Comparison of Two Means with Unknown Variances

Assumptions:
1 X 1 , . . . , X n are random samples selected from population 1
which follows a normal distribution N (µ1 , σ21 ) with
unknown variance σ21 .
2 Y1 , . . . , Ym are random samples selected from population 2
which follows a normal distribution N (µ2 , σ22 ) with
unknown variance σ22 .
3 Populations 1 and 2 are independent of each other.
Using the null hypothesis and the sample standard deviations
(unbiased) from population 1 and 2, denoted by s 1 and s 2 :
s1
1 If ∈ [0.5; 2], then assume σ21 = σ22
s2
s1
2 If ∉ [0.5; 2], then assume σ21 ̸= σ22
s2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 9 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

CASE: σ21 = σ22 = σ2

E (X − Y ) = E (X ) − E (Y ) = µ1 − µ2 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 10 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

CASE: σ21 = σ22 = σ2

E (X − Y ) = E (X ) − E (Y ) = µ1 − µ2 .

σ21 σ22
µ ¶
1 1
V (X − Y ) = V (X ) +V (Y ) = + = σ2 +
n m n m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 10 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

POOLED ESTIMATOR OF σ2

PROPOSITION 2.1
A pooled estimator of σ2 is given by the weighted
average of the two corrected sample variances S 12 and
S 22

2 (n − 1)s 12 + (m − 1)s 22
σ ≈ s p2 = (1)
n +m −2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 11 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

POOLED ESTIMATOR OF σ2

PROPOSITION 2.1
A pooled estimator of σ2 is given by the weighted
average of the two corrected sample variances S 12 and
S 22

2 (n − 1)s 12 + (m − 1)s 22
σ ≈ s p2 = (1)
n +m −2

(n − 1)s 12 + (m − 1)s 22 (n − 1) 2 (m − 1) 2
s p2 = = s1 + s =
n +m −2 n +m −2 n +m −2 2
s p2 = w s 12 + (1 − w)s 22
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 11 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 12 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2
x − y − (µ1 − µ2 )
Compute the test statistic: t 0 = r
1 1
sp +
n m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 12 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2
x − y − (µ1 − µ2 )
Compute the test statistic: t 0 = r
1 1
sp +
n m
For a significance level α, determine the rejection
region and P -value according to the rule:
Alternative Rejection Region P -value
µ1 ̸= µ2 Wα = {t 0 : |t 0 | > t α/2,n+m−2 } 2P (T > |t 0 |)
µ1 > µ2 Wα = {t 0 : t 0 > t α,n+m−2 } P (T > t 0 )
µ1 < µ2 Wα = {t 0 : t 0 < −t α,n+m−2 } P (T < t 0 )

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 12 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

EXAMPLE 2.1
A researcher claims that the stomachs of blue crabs at location A
contain more fish than those at location B . The stomach
contents of 25 blue crabs at location A were examined and
showed a sample mean of 320 mg of fish with an unbiased
standard deviation of 60 mg. Similarly, the stomach contents of
15 blue crabs at location B showed a sample mean of 280 mg of
fish with an unbiased standard deviation of 80 mg. Assume that
the amount of fish in the stomachs of blue crabs at each location
follows a normal distribution with equal variances. At a
significance level of 0.01, can the researcher’s claim be
confirmed? Can the claim be rejected?

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 13 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

s 1 60
1
= = 0.75 ∈ [0.5; 2] ⇒ σ21 = σ22 .
s 2 80

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 14 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

s 1 60
1
= = 0.75 ∈ [0.5; 2] ⇒ σ21 = σ22 .
s 2 80
2
H0 : µ1 = µ2 ; H1 : µ1 > µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 14 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

s 1 60
1
= = 0.75 ∈ [0.5; 2] ⇒ σ21 = σ22 .
s 2 80
2
H0 : µ1 = µ2 ; H1 : µ1 > µ2
2
(n − 1)s 12 + (m − 1)s 22 (25 − 1)602 + (15 − 1)802
sp = =
n +m −2 25 + 15 − 2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 14 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

s 1 60
1
= = 0.75 ∈ [0.5; 2] ⇒ σ21 = σ22 .
s 2 80
2
H0 : µ1 = µ2 ; H1 : µ1 > µ2
2
(n − 1)s 12 + (m − 1)s 22 (25 − 1)602 + (15 − 1)802
sp = =
n +m −2 25 + 15 − 2
x − y − (µ1 − µ2 ) 320 − 280 − 0
3
t0 = r = r ≈ 1.7996
1 1 1 1
sp + sp +
n m 25 15

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 14 / 33
 Comparison of Two Means with Unknown Variances Unknown but Equal Variances

STEPS OF HYPOTHESIS TESTING

s 1 60
1
= = 0.75 ∈ [0.5; 2] ⇒ σ21 = σ22 .
s 2 80
2
H0 : µ1 = µ2 ; H1 : µ1 > µ2
2
(n − 1)s 12 + (m − 1)s 22 (25 − 1)602 + (15 − 1)802
sp = =
n +m −2 25 + 15 − 2
x − y − (µ1 − µ2 ) 320 − 280 − 0
3
t0 = r = r ≈ 1.7996
1 1 1 1
sp + sp +
n m 25 15
4
t α,n+m−2 = t 0.01,38 = 2.429 > t 0 . We do not reject H0 ,
which means we cannot confirm the researcher’s
claim.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 14 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

CASE: σ21 ̸= σ22

1 When σ21 ̸= σ22 , use the test statistic

x − y − (µ1 − µ2 )
t0 = s
s 12 s 22
+
n m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 15 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

CASE: σ21 ̸= σ22

1 When σ21 ̸= σ22 , use the test statistic

x − y − (µ1 − µ2 )
t0 = s
s 12 s 22
+
n m

2 In this case, T0 follows a Student’s t -distribution with

degrees of freedom d f equal to the integer part (rounded
down) of à ! 2
s 12 s 22
+
n m
ν=
(s 12 /n)2 (s 22 /m)2
+
n −1 m −1
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 15 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

CASE: σ21 ̸= σ22

Alternative Rejection Region P -value

µ1 ̸= µ2 Wα = {t 0 : |t 0 | > t α/2,d f } 2P (T > |t 0 |)
µ1 > µ2 Wα = {t 0 : t 0 > t α,d f } P (T > t 0 )
µ1 < µ2 Wα = {t 0 : t 0 < −t α,d f } P (T < t 0 )

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 16 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

EXAMPLE 2.2
Arsenic (As) concentration (unit: ppb) in water is harmful to
health when high. Measurements of arsenic concentration were
taken in two areas: the city center and the area near the airport.
In each area, random samples were taken from 10 different
locations, and the following data were obtained:

City Center 3 7 25 10 15 6 12 25 15 7
Airport Area 48 44 40 38 33 21 20 12 1 18

Assume that arsenic concentrations are normally distributed. At

a significance level of α = 0.05, test whether there is a difference
in arsenic levels between the two areas.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 17 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

STEPS OF HYPOTHESIS TESTING

1
City Center: n = 10, x = 12.5, s 1 = 7.63.
Airport Area: m = 10, y = 27.5, s 2 = 15.35.
s1
Since = 0.49 ∉ [0.5; 2], ⇒ σ21 ̸= σ22.
s2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 18 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

STEPS OF HYPOTHESIS TESTING

1
City Center: n = 10, x = 12.5, s 1 = 7.63.
Airport Area: m = 10, y = 27.5, s 2 = 15.35.
s1
Since = 0.49 ∉ [0.5; 2], ⇒ σ21 ̸= σ22.
s2
2
H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 18 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

STEPS OF HYPOTHESIS TESTING

1
City Center: n = 10, x = 12.5, s 1 = 7.63.
Airport Area: m = 10, y = 27.5, s 2 = 15.35.
s1
Since = 0.49 ∉ [0.5; 2], ⇒ σ21 ̸= σ22.
s2
2
H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
3
Use the test statistic
x − y − (µ1 − µ2 ) x−y
t0 = s =s ≈ −2.77
2 2 2 2
s1 s2 s1 s2
+ +
n m n m
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 18 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

STEPS OF HYPOTHESIS TESTING

1
City Center: n = 10, x = 12.5, s 1 = 7.63.
Airport Area: m = 10, y = 27.5, s 2 = 15.35.
s1
Since = 0.49 ∉ [0.5; 2], ⇒ σ21 ̸= σ22.
s2
2
H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
3
Use the test statistic
x − y − (µ1 − µ2 ) x−y
t0 = s =s ≈ −2.77
2 2 2 2
s1 s2 s1 s2
+ +
n m n m
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 18 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

1
In this case, T0 follows a Student’s t -distribution
with degrees of freedom d f equal to the integer
part (rounded down) of
¶2
s 12 s 22
µ
+
n m
ν= 2 = 13.19 ⇒ d f = 13
(s 1 /n)2 (s 22 /m)2
+
n −1 m −1

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 19 / 33
 Comparison of Two Means with Unknown Variances Unknown and Unequal Variances

1
In this case, T0 follows a Student’s t -distribution
with degrees of freedom d f equal to the integer
part (rounded down) of
¶2
s 12 s 22
µ
+
n m
ν= 2 = 13.19 ⇒ d f = 13
(s 1 /n)2 (s 22 /m)2
+
n −1 m −1

2
At α = 0.05, we have t α/2,d f = t 0.025,13 = 2.160. Since
|t 0 | = 2.77 > t 0.025,13 = 2.160, we reject the null
hypothesis H0 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 19 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION

WITH L ARGE S AMPLE S IZES

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 with any distribution, having mean
µ1 and variance σ21 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 20 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION

WITH L ARGE S AMPLE S IZES

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 with any distribution, having mean
µ1 and variance σ21 .
2
Y1 , . . . , Ym are random samples selected from
population 2 with any distribution, having mean
µ2 and variance σ22 .

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 20 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION

WITH L ARGE S AMPLE S IZES

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 with any distribution, having mean
µ1 and variance σ21 .
2
Y1 , . . . , Ym are random samples selected from
population 2 with any distribution, having mean
µ2 and variance σ22 .
3
Populations 1 and 2 (represented by X and Y ) are
independent of each other.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 20 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

COMPARISON OF TWO MEANS FOR ANY DISTRIBUTION

WITH L ARGE S AMPLE S IZES

Assumptions:
1
X 1 , . . . , X n are random samples selected from
population 1 with any distribution, having mean
µ1 and variance σ21 .
2
Y1 , . . . , Ym are random samples selected from
population 2 with any distribution, having mean
µ2 and variance σ22 .
3
Populations 1 and 2 (represented by X and Y ) are
independent of each other.
4
Sample sizes are sufficiently large.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 20 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 21 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2
x − y − (µ1 − µ2 )
Compute: z 0 = s
s 12 s 22
+
n m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 21 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

STEPS OF HYPOTHESIS TESTING

H0 : µ1 = µ2
x − y − (µ1 − µ2 )
Compute: z 0 = s
s 12 s 22
+
n m
For a given significance level α
Alternative Rejection Region P -value
µ1 ̸= µ2 Wα = {z 0 : |z 0 | > z α/2 } 2[1 − Φ(|z 0 |)]
µ1 > µ2 Wα = {z 0 : z 0 > z α } 1 − Φ(z 0 )
µ1 < µ2 Wα = {z 0 : z 0 < −z α } Φ(z 0 )

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 21 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

EXAMPLE 3.1
A survey on the height of students from two faculties:
Mathematics and Information Technology (IT). A
random sample of 50 students from the Mathematics
faculty has a sample mean height of 163 cm and an
unbiased standard deviation of 5 cm. Another sample
of 50 students from the IT faculty has a sample mean
height of 166 cm and an unbiased standard deviation
of 8 cm. At a significance level of α = 1%, determine
whether there is a difference in the average height of
students between the two faculties.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 22 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

STEPS OF HYPOTHESIS TESTING

1
n = m = 50 are sufficiently large,
x = 163, s 1 = 5; y = 166, s 2 = 8;
2
H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 23 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

STEPS OF HYPOTHESIS TESTING

1
n = m = 50 are sufficiently large,
x = 163, s 1 = 5; y = 166, s 2 = 8;
2
H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
x − y − (µ1 − µ2 ) 163 − 166 − 0
3
z0 = s = s ≈ −2.25
s 12 s 22 52 82
+ +
n m 50 50

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 23 / 33
 Comparison of Two Means for Any Distribution with Large
Sample Sizes

STEPS OF HYPOTHESIS TESTING

1
n = m = 50 are sufficiently large,
x = 163, s 1 = 5; y = 166, s 2 = 8;
2
H0 : µ1 = µ2 ; H1 : µ1 ̸= µ2
x − y − (µ1 − µ2 ) 163 − 166 − 0
3
z0 = s = s ≈ −2.25
s 12 s 22 52 82
+ +
n m 50 50
4
With α = 0.01 ⇒ z α/2 = z 0.005 ≈ 2.58.
Therefore, since |z 0| < z α/2, we do not
have enough evidence to reject H0.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 23 / 33
 Comparison of Two Population Proportions

COMPARISON OF TWO POPULATION PROPORTIONS

Investigate the proportion of elements satisfying

a certain property A in two independent
populations, with corresponding proportions p 1
and p 2 .
From the two populations, select two random
samples with sufficiently large sample sizes, n
and m , respectively. Let X and Y be the number
of elements satisfying property A in sample 1 and
sample 2. Then X ∼ B (n, p 1 ) and Y ∼ B (m, p 2 ).

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 24 / 33
 Comparison of Two Population Proportions

COMPARISON OF TWO POPULATION PROPORTIONS

Investigate the proportion of elements satisfying

a certain property A in two independent
populations, with corresponding proportions p 1
and p 2 .
From the two populations, select two random
samples with sufficiently large sample sizes, n
and m , respectively. Let X and Y be the number
of elements satisfying property A in sample 1 and
sample 2. Then X ∼ B (n, p 1 ) and Y ∼ B (m, p 2 ).
We aim to compare the two proportions p 1 and
p 2.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 24 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

H0 : p 1 = p 2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 25 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

H0 : p 1 = p 2
Compute the test statistic:
pˆ1 − pˆ2 − (p 1 − p 2 )
z0 = s µ ¶
1 1
p̂(1 − p̂) +
n m

x y x+y
where pˆ1 = , pˆ2 = , and p̂ =
n m n +m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 25 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

H0 : p 1 = p 2
Compute the test statistic:
pˆ1 − pˆ2 − (p 1 − p 2 )
z0 = s µ ¶
1 1
p̂(1 − p̂) +
n m

xy x+y
where pˆ1 = , pˆ2 = , and p̂ =
nm n +m
For significance level α, determine:
Alternative Rejection Region P -value
p 1 ̸= p 2 Wα = {z 0 : |z 0 | > z α/2 } 2[1 − Φ(|z 0 |)]
p1 > p2 Wα = {z 0 : z 0 > z α } 1 − Φ(z 0 )
p1 < p2 Wα = {z 0 : z 0 < −z α } Φ(z 0 )

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 25 / 33
 Comparison of Two Population Proportions

EXAMPLE 4.1
A pharmaceutical company wants to test whether a
new drug reduces the occurrence of chest pain in
patients. The company conducts an experiment on
400 patients who previously experienced chest pain,
dividing them into two groups: group 1 includes 200
people who received the drug, and group 2 includes
200 people who received a placebo. After monitoring,
only 8 people in group 1 and 25 people in group 2
continued to experience chest pain. At a significance
level of α = 0.05, determine whether there is a
difference in the proportion of patients still
experiencing chest pain after taking the drug.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 26 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

8 25 x+y 8 + 25
1
pˆ1 = and pˆ2 = ; p̂ = = ;
200 200 n + m 200 + 200
n = m = 200.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 27 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

8 25 x+y 8 + 25
1
pˆ1 = and pˆ2 = ; p̂ = = ;
200 200 n + m 200 + 200
n = m = 200.
2
H0 : p 1 = p 2 ; H1 : p 1 ̸= p 2

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 27 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

8 25 x+y 8 + 25
1
pˆ1 = and pˆ2 = ; p̂ = = ;
200 200 n + m 200 + 200
n = m = 200.
2
H0 : p 1 = p 2 ; H1 : p 1 ̸= p 2
3

pˆ1 − pˆ2 − (p 1 − p 2 ) pˆ1 − pˆ2

z0 = s µ ¶ = s µ ¶ ≈ −3.09
1 1 1 1
p̂(1 − p̂) + p̂(1 − p̂) +
n m n m

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 27 / 33
 Comparison of Two Population Proportions

STEPS OF HYPOTHESIS TESTING

8 25 x+y 8 + 25
1
pˆ1 = and pˆ2 = ; p̂ = = ;
200 200 n + m 200 + 200
n = m = 200.
2
H0 : p 1 = p 2 ; H1 : p 1 ̸= p 2
3

pˆ1 − pˆ2 − (p 1 − p 2 ) pˆ1 − pˆ2

z0 = s µ ¶ = s µ ¶ ≈ −3.09
1 1 1 1
p̂(1 − p̂) + p̂(1 − p̂) +
n m n m

4
At α = 0.05 ⇒ z α/2 = z 0.025 ≈ 1.96. Since |z 0 | > z α/2 , we
reject H0 .
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 27 / 33
 Comparison of Two Population Proportions

LESSON SUMMARY

1 If two independent samples follow a normal distribution

and the population variances σ21 and σ22 are known, then
use the standard normal distribution (instead of the
Student’s t -distribution) for hypothesis testing.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 28 / 33
 Comparison of Two Population Proportions

LESSON SUMMARY

1 If two independent samples follow a normal distribution

and the population variances σ21 and σ22 are known, then
use the standard normal distribution (instead of the
Student’s t -distribution) for hypothesis testing.
2 In the case of small sample sizes with unknown population
variances σ21 and σ22 from two independent normal
populations, use the Student’s t -distribution for testing.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 28 / 33
 Comparison of Two Population Proportions

LESSON SUMMARY

1 If two independent samples follow a normal distribution

and the population variances σ21 and σ22 are known, then
use the standard normal distribution (instead of the
Student’s t -distribution) for hypothesis testing.
2 In the case of small sample sizes with unknown population
variances σ21 and σ22 from two independent normal
populations, use the Student’s t -distribution for testing.
3 For large samples, if the population variances σ21 and σ22 are
known, use them in the test. If not, use the sample
variances s 12 and s 22 to perform the test.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 28 / 33
 Paired Sample Testing

1
Let (X 1, Y1), (X 2, Y2), . . . , (X n , Yn ) be a set of n
observed data pairs, where X is a
population with mean µ1 and variance
σ21 , and Y is a population with mean µ2
and variance σ22.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 29 / 33
 Paired Sample Testing

1
Let (X 1, Y1), (X 2, Y2), . . . , (X n , Yn ) be a set of n
observed data pairs, where X is a
population with mean µ1 and variance
σ21 , and Y is a population with mean µ2
and variance σ22.
2
X and Y are dependent (paired samples).

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 29 / 33
 Paired Sample Testing

1
Let (X 1, Y1), (X 2, Y2), . . . , (X n , Yn ) be a set of n
observed data pairs, where X is a
population with mean µ1 and variance
σ21 , and Y is a population with mean µ2
and variance σ22.
2
X and Y are dependent (paired samples).
3
Assume D = X − Y follows a normal
distribution with mean
µD = E (X − Y ) = E (X ) − E (Y ) = µ1 − µ2
and the population variance σ2D is
unknown.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 29 / 33
 Paired Sample Testing

STEPS OF HYPOTHESIS TESTING

H0 : µD = µ1 − µ2 = ∆0

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 30 / 33
 Paired Sample Testing

STEPS OF HYPOTHESIS TESTING

H0 : µD = µ1 − µ2 = ∆0
d
Compute the test statistic: t 0 = p
sD / n

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 30 / 33
 Paired Sample Testing

STEPS OF HYPOTHESIS TESTING

H0 : µD = µ1 − µ2 = ∆0
d
Compute the test statistic: t 0 = p where d is
sD / n
the sample mean of the n differences
D 1 , D 2 , . . . , D n , and s D is the standard deviation of
these differences.
For a given significance level α, determine the
rejection region and the P -value.
Alternative Rejection Region P -value
µD ̸= ∆0 Wα = {t 0 : |t 0 | > t α/2,n−1 } 2P (T > |t 0 |)
µD > ∆0 Wα = {t 0 : t 0 > t α,n−1 } P (T > t 0 )
µD < ∆0 Wα = {t 0 : t 0 < −t α,n−1 } P (T < t 0 )
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 30 / 33
 Paired Sample Testing

EXAMPLE 5.1
The concentration of metals in drinking water affects water
quality, and abnormally high metal levels may pose health risks.
Zinc concentration (in mg/L) was measured at the surface and
bottom of a river at 6 locations, as shown in the table below:

Zinc Concentration 1 2 3 4 5 6
River Bottom 0.430 0.266 0.567 0.531 0.707 0.716
Surface 0.415 0.238 0.390 0.410 0.605 0.609
Difference 0.015 0.028 0.177 0.121 0.102 0.107

Do the data provide evidence, at the 0.05 significance level, that

the average zinc concentration at the bottom of the river is
higher than at the surface?

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 31 / 33
 Paired Sample Testing

Let D be the difference in zinc concentration

between the bottom and surface at the same
sampling location. The sample mean of D is
d = 0.0917 and the unbiased sample standard
deviation is s D = 0.0607.

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 32 / 33
 Paired Sample Testing

Let D be the difference in zinc concentration

between the bottom and surface at the same
sampling location. The sample mean of D is
d = 0.0917 and the unbiased sample standard
deviation is s D = 0.0607.
d 0.0917
t0 = p = p = 3.7005
s D / n 0.0607/ 6

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 32 / 33
 Paired Sample Testing

Let D be the difference in zinc concentration

between the bottom and surface at the same
sampling location. The sample mean of D is
d = 0.0917 and the unbiased sample standard
deviation is s D = 0.0607.
d 0.0917
t0 = p = p = 3.7005
s D / n 0.0607/ 6
At a significance level of α = 0.05, we have
t α,n−1 = t 0.05,5 = 2.0150. Since t 0 > t α,n−1 , we reject H0 .
Therefore, there is sufficient evidence to
conclude that the zinc concentration at the
bottom of the river is higher than at the surface.
Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 32 / 33
 Paired Sample Testing

THANK YOU ALL FOR YOUR ATTENTION

Dr. Lê Xuân Đại (HCMUT-OISP) STATISTICAL INFERENCE FOR TWO SAMPLES HCMC — 2025. 33 / 33