Scribd
Upload
218 views

Hypothesis Testing For The Difference of Proportions

This document discusses hypothesis testing for the difference between two population proportions. It outlines how to formulate the null and alternative hypotheses for two-tailed, upper-tailed, and lower-tailed tests when comparing the difference between two population proportions. The test statistic used is a z-score calculated based on the sample proportions, population proportions, sample sizes, and the proportion of the combined sample. An example compares the effectiveness of a new and old cholesterol drug and conducts a hypothesis test to determine if the new drug is more effective than the old at the 99% confidence level based on samples from people taking each drug. The results do not reject the null hypothesis that there is no difference in effectiveness.

Uploaded by

AmarnathMaiti
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
218 views

Hypothesis Testing For The Difference of Proportions

This document discusses hypothesis testing for the difference between two population proportions. It outlines how to formulate the null and alternative hypotheses for two-tailed, upper-tailed, and lower-tailed tests when comparing the difference between two population proportions. The test statistic used is a z-score calculated based on the sample proportions, population proportions, sample sizes, and the proportion of the combined sample. An example compares the effectiveness of a new and old cholesterol drug and conducts a hypothesis test to determine if the new drug is more effective than the old at the 99% confidence level based on samples from people taking each drug. The results do not reject the null hypothesis that there is no difference in effectiveness.

Uploaded by

AmarnathMaiti
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Hypothesis testing for the difference of

proportions
Now that we know how to find a confidence interval around the difference
of proportions, let’s look at how to conduct a hypothesis test with the
difference of proportions, when we want to use the difference of sample
proportions to make an inference about the difference of population
proportions.

Building hypothesis statements


The null and alternative hypotheses will always be formulated in terms of
the difference between the two population proportions, p1 − p2, and we can
have three different scenarios.

In a two-tailed test, the null hypothesis will state that the proportions don’t
differ, whereas the alternative hypothesis states that there is a difference
between proportions. So we write the hypothesis statements for a two-
tailed test as

H0 : p1 − p2 = 0

Ha : p1 − p2 ≠ 0

or

H0 : p1 = p2

Ha : p1 ≠ p2

347
In an upper-tailed test, the alternative hypothesis states that the
difference in proportions is positive, so we write

H0 : p1 − p2 ≤ 0

Ha : p1 − p2 > 0

or

H0 : p1 ≤ p2

Ha : p1 > p2

In a lower-tailed test, the alternative hypothesis states that the difference


in means is negative, so we write

H0 : p1 − p2 ≥ 0

Ha : p1 − p2 < 0

or

H0 : p1 ≥ p2

Ha : p1 < p2

Calculating the test statistic


As long as we take independent random samples from each population,
and n1 p1̂ ≥ 5, n1(1 − p1̂ ) ≥ 5, n2 p2̂ ≥ 5, and n2(1 − p2̂ ) ≥ 5, then the test statistic
formula we’ll use is

348
( p1̂ − p2̂ ) − (p1 − p2)
z=
̂ − p)̂ ( n1 +
p(1 n2 )
1
1

where p1̂ and p2̂ are the sample proportions, p1 and p2 are the population
proportions, n1 and n2 are the sample sizes, and p̂ is the proportion of the
combined sample, given by

p1̂ n1 + p2̂ n2
p̂ =
n1 + n2

which we can also write as

x1 + x2
p̂ =
n1 + n2

where x1 and x2 are the number of “successes” in each sample. We say


that the null hypothesis always states a zero difference between
population proportions, such that p1 − p2 = 0, so the test statistic formula
actually simplifies to

p1̂ − p2̂
z=
̂ − p)̂ ( n1 +
p(1 n2 )
1
1

Let’s rework the same example from the previous section.

Example

A team of scientists claims that a new cholesterol lowering drug is more


affective than an older version. The team takes two random samples of 250

349
people, and for 3 months administer the new drug to the first group and
the old drug to the second group. 120 people in the first group and and 107
people in the second group show decreased cholesterol levels. Can the
team conclude at a 99 % confidence level that the new drug is more
affective than the old drug at lowering cholesterol?

If p1 is the proportion of population 1 (the population that takes the new


drug) whose cholesterol decreases, and p2 is the proportion of population
2 (the population that takes the old drug) whose cholesterol decreases,
then the null and alternative hypotheses are

H0 : p1 − p2 ≤ 0

Ha : p1 − p2 > 0

The pooled proportion is

x1 + x2
p̂ =
n1 + n2

120 + 107
p̂ =
250 + 250

227
p̂ =
500

p̂ = 0.454

and the sample proportions are

350
120
p1̂ = = 0.480
250

107
p2̂ = = 0.428
250

So the test statistic is

p1̂ − p2̂
z=
̂ − p)̂ ( n1 +
p(1 n2 )
1
1

0.480 − 0.428
z=
0.454(1 − 0.454)( 250 250 )
1 1
+

0.052
z=
0.454(0.546)( 125 )
1

0.052
z=
0.247884
125

125
z = 0.052
0.247884

z ≈ 1.17

Now we need to determine the critical z-value. Our level of significance is


α = 0.01, and for a right-tailed test the corresponding z-value is z = 2.33.

351
Using the critical-value approach, we’ll therefore reject H0 if z ≥ 2.33. Since
1.17 ≱ 2.33, the team can’t reject the null hypothesis. Therefore, they’re
unable to provide support for the hypothesis that the new drug is more
affective than the old drug.

352

You might also like