Hypothesis Testing with Two

Samples
Claims are often studied using samples from two groups.

 A company wants to know if their new windshield treatment repels

water more effectively than an untreated windshield.
 It is speculated that teenagers sleep more than adults on average.

In this chapter we will:

 Apply hypothesis tests to compare means of two independent groups
 Apply hypothesis tests to compare means of two dependent groups
(matched pairs)
 Apply hypothesis tests to compare two proportions
 Find confidence intervals of two independent groups, matched pairs,
and two proportions
 Apply hypothesis tests of variance
Symbols/Formulas Needed
sample mean
population mean
population standard deviation
sample standard deviation
̂ sample proportion of success
population proportion of success
population proportion of failure
sample size
Test statistic for
comparison of means
of large independent
groups
Test statistic for
comparison of means
of small independent
groups
Margin of error for
large independent ⁄ ∙
groups
Margin of error for
small independent ⁄ ∙
groups
Test statistic for
comparison of
dependent groups √
(matched pairs)
Margin of error for ∙
⁄
matched pairs √
Test statistic for ̂ ̂
comparison of
proportions
Margin of error for
̂ ̂
proportions ⁄ ∙
Types of Groups:
Identifying the relationship of the two groups is the first step in a test with
two samples.

independent groups – the two sample groups are not related as the
outcome of one does not influence the other

matched pairs – subjects in the two sample groups have a natural pairing,
often they are actually the same subjects

Example:
Are these independent groups or matched pairs?

a) A company wants to know if their new windshield treatment repels

water more effectively than an untreated windshield. Ten windshields
were tested by simulating rain. The same ten windshields were then
treated and the experiment performed again.

matched pairs

b) It is speculated that teenagers sleep more than adults on average.

To study this, a sample of 16 teens had a mean of 8.9 hours slept per
night with a sample standard deviation of 1.2 hours. A sample of 12
adults had a mean of 6.9 hours per night with a sample standard
deviation of 0.6 hours.

independent groups

The relationship of the groups changes how the test statistic is calculated.

Recall the steps to perform a hypothesis test:

1) State the null and alternative hypotheses.
2) Calculate the test statistic from the sample data.
3) Find the p-value.
4) Compare the p-value to the significance level.
5) Decide to either reject the null hypothesis or fail to reject the null
hypothesis.
Two Independent Groups:
We always want to know if there is a difference between the populations
being studied. So, this test is the same as for one sample but uses the
difference of each measure in its calculation of the test statistic.

Here are the new formulas for the test statistic.

or

When using the t-score, take the degrees of freedom to be the smaller of
1 and 1.

Example:
It is speculated that teenagers sleep more than adults on average. To
study this, a sample of 16 teens had a mean of 8.9 hours slept per night
with a sample standard deviation of 1.2 hours. A sample of 12 adults had a
mean of 6.9 hours per night with a sample standard deviation of 0.6 hours.
These populations are normally distributed.
At a significance level of 0.05, is there enough evidence that teenagers
sleep more than adults per night on average?

: 0
: 0 (right-tailed test)

8.9 1.2 16
6.9 0.6 12

8.9 6.9 0
5.77
1.2 0.6
16 12

5.77 with 11
0.00006

0.05 is greater than 0.00006

Reject the null hypothesis

 There is enough evidence that teenagers sleep more than adults per
night on average.

Problem #1:
The results of a study on body temperatures of men and women are shown
in the table. The data are from random samples from normally distributed
populations.

Men ( ) Women ( )
97.57 97.46
0.84 0.65
45 59

Using a 0.10 significance level, test the claim that men have a different
body temperature than women.
For confidence intervals, use the differences of measures to find the point
estimate and the margin of error.
The point estimate is

and the margin of error is

⁄ ∙ or ⁄ ∙

Example:
From the data on men's and women's body temperatures, find the 95%
confidence interval for the claim that men have a higher body temperature
than women. Round to 2 decimal places.

point estimate 97.57 97.46 0.11

95% confidence → ⁄ 1.96

. .
1.96 0.30

0.11 0.30, 0.11 0.30

0.19, 0.41

Matched Pairs:
To find the test statistic, the differences in each of the pairs are used to
calculate the mean difference and the standard deviation of differences.
Only the t-score applies in this context.

To find the test statistic, calculate

 √

where is the mean of the differences and is the standard deviation of

the differences. To find these values, first calculate the difference of each
pair.

Example:
A college athletic trainer believes by taking a special vitamin weight lifters
can increase strength. Eight athletes are randomly selected and given a
test of strength, using the standard bench press. After 2 weeks of
supplementation with the vitamin, the athletes are tested again.

Athlete 1 2 3 4 5 6 7 8
Before ( ) 210 230 180 205 260 255 220 215

After ( ) 220 235 180 200 270 250 225 215

Test the claim that the vitamin will increase strength at 0.05.

: 0
: 0 (left-tailed test)
For the test statistic, first list all of the differences.

210 220 10 10
230 235 5 5
0 5
5 0
2.5
5.976143

2.5 0
1.18
5.976143
√8
 1.18 with 7
0.1383

0.1383 is greater than 0.05

Fail to reject the null hypothesis.

There is not enough evidence that this vitamin will increase strength.

Problem #2:

A high school athletic coach believes by following a particular training

program athletes will lower their 800 meter run time. Seven athletes are
randomly selected and timed. After 4 weeks of the program, the athletes
are timed again. Times are given in seconds.

Athlete 1 2 3 4 5 6 7
Before ( ) 130 133 122 128 141 136 116

After ( ) 122 123 115 130 130 132 114

Test the claim that the program will lower the 800 meter run time at
0.075.
The confidence interval for matched pairs uses point estimate

and margin of error given by

⁄ ∙
√

Problem #3:

Find the 90% confidence interval for the high school 800 meter runs times.
Two Proportions:
The two proportion samples are always from independent groups. The test
statistic is

̂ ̂

Here is the total pooled proportion. It is found by the total number of

successes from both samples divided by the total number of trials.

Also, ̂ and ̂ are the proportions of success of the two samples.

Example:
In a study on nursing home care, a researcher found that 12 out of 34 small
nursing home had an activities director employed at the home, while 17 out
of 24 large nursing homes had an activities director.

At 0.05, test the claim that there is a difference in the proportions of the
small and large nursing homes with an activities director.
: 0
: 0 (two-tailed test)

̂ 0.353 34 ̂ 0.708 24

0.5 0.5

0.353 0.708 0
2.66
0.5 0.5 0.5 0.5
34 24

2∙ 2.66
2 0.0039 0.0078

0.05 is greater than 0.0078

 Reject the null hypothesis.

There is enough evidence that there is a difference in the proportions

of the small and large nursing homes with an activities director.

Problem #4:

Investors want to know how the Manhattan apartment market compares to

the San Francisco apartment market over the previous 6 months. From
Manhattan, a random sample of 85 apartments showed that 66 had
increased in value. In San Francisco, a random sample of 170 apartments
had 110 that increased in value.

The investors wish to determine if the proportion of apartments in each city

that increased is different using 0.10. Conduct a hypothesis test to find
out.
The confidence interval for two proportions has point estimate
̂ ̂
and margin of error

̂ ̂
⁄ ∙

Problem #5:

Find the 99% confidence interval for the Manhattan and San Francisco
apartment data.
Wrap Up
This chapter has eight complex formulas for test statistics and margin of
error.
Do you notice any patterns among the four test statistic formulas?
Do you notice any patterns among the four margin of error formulas?
Is there anything that connects each test statistic formula and its
corresponding margin of error formula?
Are there any clear differences or unusual features among these formulas?