LESSON 12

INFERENCES ABOUT TWO PROPORTIONS

LEARNING OBJECTIVE:
• Conduct a formal hypothesis test of a claim about two population proportions (p).

Assumptions:

1. We have two proportions from two independent simple random samples.

2. For both samples, the conditions np  5 and nq  5 are satisfied. That is, there are at least five successes
and five failures in each of the two samples.

The following test statistic applies to null and alternative hypotheses that fit one of these three formats:

Two-Tailed One-Tailed (Left) One-Tailed (Right)

H0: p1 = p2 H0: p1 ≥ p2 H0: p1 ≤ p2
H1: p1  p2 H1: p1 < p2 H1: p1 > p2

Notation for Two Proportions

Population 1 Population 2

p1 = population proportion p2 = population proportion

n1 = size of sample n2 = size of sample
x1 = number of successes in the sample x2 = number of successes in the sample
x x
pˆ 1 = 1 (the sample proportion) pˆ 2 = 2 (the sample proportion)
n1 n2
qˆ1 = 1 − pˆ 1 qˆ 2 = 1 − pˆ 2

Pooled Estimate of p1 and p2

The pooled estimate of p1 and p2 is denoted by p and is given by
x + x2
p= 1
n1 + n 2
We denote the complement of p by q , so q = 1 − p

Test Statistic
𝑝̂1 − 𝑝̂ 2 𝑥1 𝑥2 𝑥1 + 𝑥2
𝑧= 𝑤ℎ𝑒𝑟𝑒 𝑝̂1 = , 𝑝̂ 2 = , 𝑝̅ = , 𝑞̅ = 1 − 𝑝̅
𝑝̅ 𝑞̅ 𝑝̅ 𝑞̅ 𝑛1 𝑛2 𝑛1 + 𝑛2
√ +
𝑛1 𝑛2

Note: When H0 is rejected, the conclusion is the statement in H1

When H0 is not rejected, the conclusion is the statement in H0

A Self-regulated Learning Module 1

Example 1: Would you favor spending more money on the gadgets? This question was asked by a research group.
Of a random sample of 220 women, 59 responded “yes”. Another random sample of 175 men showed that 56
responded “yes”. Does this information indicate a difference (either way) between the population proportion of
women and the population proportion of men who favor spending more money on gadgets? Use  = 5%.

SOLUTION

State the null and alternative hypotheses

H0: There is no difference between the proportion of women and proportion of men who favor spending more
money on gadgets. (p1 = p2)

H1: There is a difference between the proportion of women and proportion of men who favor spending more
money on gadgets. (p1  p2)

Data:
Women Men
n1 = 220 n2 = 175
x1 = 59 x2 = 56
59 56
𝑝̂1 = = 0.27 𝑝̂ 2 = = 0.32
220 175

Pooled Estimate for p1 and p2

59 + 56 115
𝑝̅ = = = 0.29 𝑞̅ = 1 − 𝑝̅ = 1 − 0.29 = 0.71
220 + 175 395

Test Statistic

𝑝̂1 − 𝑝̂2 0.27 − 0.32 −0.05

𝑧= = = = −1.094
𝑝̅ 𝑞̅ 𝑝̅𝑞̅ 0.0457
√ √(0.29)(0.71) + (0.29)(0.71)
𝑛1 + 𝑛2 225 175

A Self-regulated Learning Module 2

Critical Value(s)
 = 5% = 0.05
H1: (p1  p2)  indicates a two-tailed test, thus, the critical values are positive and negative

Critical Region Non Critical Critical Region

Reject H0 Region Reject H0
Do not Reject H0

– 1.96 – 1.094 1.96

Decision: Do not reject H0 since the computed z-value of – 1.094 is not within the critical region

Conclusion: At 5% level, there is no difference between the proportion of women and proportion of men who
favor spending more money on gadgets.

Research Data Result Interpretation:

The proportion of women who favor spending more money on the gadgets is not statistically significant (– .05)
than the proportion of men who favor spending more money on gadgets, z = – 1.094, p > 0.05.

Example 2: A random sample of 153 people ages 16 to 19 was taken from Benguet, and 12 were found to be
high school dropouts. Another random sample of 128 people ages 16 to 19 was taken from Apayao and 7 were
found to be high school dropouts. Do these data indicate that the population proportion of high school dropouts
on Benguet is different (either way) from that of Apayao? Use a 1% level of significance.

SOLUTION:

State the null and alternative hypotheses

H0: The proportion of high school dropouts on Benguet and Apayao is the same. (p1 = p2)
H1: The proportion of high school dropouts on Benguet is different from that of Apayao (p1  p2)

Data:
Benguet Apayao
n1 = 153 n2 = 128
x1 = 12 x2 = 7
12 7
𝑝̂1 = = 0.08 𝑝̂ 2 = = 0.06
153 128

A Self-regulated Learning Module 3

Pooled Estimate for p1 and p2

12 + 7 19
𝑝̅ = = = 0.07 𝑞̅ = 1 − 𝑝̅ = 1 − 0.07 = 0.93
153 + 128 281

Test Statistic
𝑝̂1 − 𝑝̂ 2 0.08 − 0.06 0.02
𝑧= = = = 0.654
𝑝̅𝑞̅ 𝑝̅𝑞̅ 0.0306
√ √(0.07)(0.93) + (0.07)(0.93)
𝑛1 + 𝑛2 153 128

Critical Value(s)
 =1% = 0.01
H1: (p1  p2)  indicates a two-tailed test

Critical Region Non Critical Critical Region

Reject H0 Region Reject H0
Do not Reject H0

– 2.58 0.654 2.58

Decision: Do not reject H0 since the computed z-value of 0.654 is not within the critical region

Conclusion: At 1% level, the proportion of high school dropouts in Benguet and Apayao is the same.

Research Data Result Interpretation:

The proportion of high school dropouts in Benguet is not statistically significant (.20) than the proportion of high
school dropouts in Apayao, z = 0.654, p > 0.05.

A Self-regulated Learning Module 4

Example 3: A random sample of 228 voters registered in Baguio City showed that 141 voted in the last general
election. A random sample of 216 registered voters in Itogon, Benguet showed that 125 voted in the most recent
general election. Do these data indicate that the population proportion of voter turnout in Baguio City is different
from Itogon Benguet? Use a 5% level of significance.

SOLUTION:

State the null and alternative hypotheses

H0: The proportion of voter turnout in Baguio City and in Itogon Benguet is the same. (p1 = p2)
H1: The proportion of voter turnout in Baguio City is different from Itogon Benguet. (p1  p2)

Data:
Baguio Itogon
n1 = 228 n2 = 216
x1 = 141 x2 = 125
141 125
𝑝̂1 = = 0.62 𝑝̂ 2 = = 0.58
228 216

Pooled Estimate for p1 and p2

141 + 125 266

𝑝̅ = = = 0.60 𝑞̅ = 1 − 𝑝̅ = 1 − 0.60 = 0.40
228 + 216 444

Test Statistic

𝑝̂1 − 𝑝̂2 0.62 − 0.58 1.20

𝑧= = = = 25.81
𝑝̅ 𝑞̅ 𝑝̅𝑞̅ 0.0465
√ √(0.60)(0.40) + (0.60)(0.40)
𝑛1 + 𝑛2 228 216

Critical Values
 =5% = 0.05
H1: (p1  p2)  indicates a two-tailed test

A Self-regulated Learning Module 5

 Critical Region Non Critical Critical Region
Reject H0 Region Reject H0
Do not Reject H0

– 1.96 1.96 25.81

Decision: Reject H0 since the computed z-value of 25.81 is within the critical region

Conclusion: At 5& level, the proportion of voter turnout in Baguio City is different from Itogon Benguet. In fact
proportion of voter turnout in Baguio City is significantly higher compared to Itogon Benguet.

Research Data Result Interpretation:

The proportion of voter turnout in Baguio City is statistically significant (1.20) than the proportion of voter turnout
in Itogon, Benguet, z = 25.81, p < 0.05.

Example 4: Among 198 smokers who underwent a “sustained care” program, 51 were no longer smoking after
six months. Among 199 smokers who underwent a “standard care” program, 30 were no longer smoking after six
months. Use 1% significance level to test the claim that the rate of success for smoking cessation is different with
the sustained care program compared to standard care program.

SOLUTION:

State the null and alternative hypotheses

H0: The rate of success for smoking cessation is the same with the sustained care program compared to
standard care program. (p1 = p2)
H1: The rate of success for smoking cessation is different with the sustained care program compared to
standard care program. (p1 ≠ p2)

Sustained Care Standard Care

n1 = 198 n2 = 199
x1 = 51 x2 = 30
51 30
𝑝̂1 = = 0.258 𝑝̂2 = = 0.151
198 199

Pooled Estimate for p1 and p2

51 + 30 81
𝑝̅ = = = 0.204 𝑞̅ = 1 − 𝑝̅ = 1 − 0.204 = 0.796
198 + 199 397

Test Statistic
𝑝̂1 − 𝑝̂ 2 0.258 − 0.151 0.107
𝑧= = = = 2.675
𝑝̅𝑞̅ 𝑝̅𝑞̅ 0.040
√ √(0.204)(0.796) + (0.204)(0.796)
𝑛1 + 𝑛2 198 199

A Self-regulated Learning Module 6

Critical Values
 =1% = 0.01
H1: (p1  p2)  indicates a two-tailed test

Critical Region Non Critical Critical Region

Reject H0 Region Reject H0
Do not Reject H0

– 2.58 2.58 2.675

Decision: Reject H0 since the computed z-value of 2.675 is within the critical region

Conclusion: At 1% level, the rate of success for smoking cessation is different with the sustained care program
compared to standard care program.

Research Data Result Interpretation:

The rate of success for smoking cessation with the sustained care program is statistically significant (.107) than
the rate of success with standard care program, z = 2.675, p < 0.01.

EXERCISES

1. Generally speaking, would you say that most people can be trusted? A random sample of 250 people in Baguio
City ages 18–25 showed that 45 said yes. Another random sample of 280 people in Baguio City ages 35–45
showed that 71 said yes. Does this indicate that the population proportion of trusting people in Baguio is
different for the older group compared to the younger group? Use  = 0.05.

2. Out of a random sample of 100 adult Filipinos from urban communities, 37 people believe in miracles.
However, out of a random sample of 100 adult Filipinos from the rural communities, 47 claims that they
believe in miracles. Does this indicate that the proportion of people from urban communities who believed in
miracles is different from the proportion of people in rural communities? Use  = 1%.

3. In a study of JoyfulBee drive-through orders, it was found that 264 orders were accurate and 54 were not
accurate. For MangDonald’s, 329 orders were found to be accurate while 33 orders were not accurate. Use a
0.05 significance level to test the claim that JoyfulBee and MangDonald’s have the same accuracy rates.

A Self-regulated Learning Module 7