Section 10.

1: Hypothesis Testing for Two Population Proportions (p1 and p2)

1. Define the two parameters in words:

p1 = proportion of the first population that ______________
p2 = proportion of the second population that ______________
Specify . If  is not given, pick a value for it. Typical values include 5%, 1% and 10%.

2. State your hypotheses.

Null Hypothesis:
Ho: p1 = p2

Alternative Hypothesis (Pick one):

Ha: p1 ≠ p2 [Two-tailed test]
Ha: p1 > p2 [Upper tail test]
Ha: p1 < p2 [Lower tail test]

3. Collect your 2 random samples and calculate the test statistic

n1 independent trials from population 1
x1 = total number of successes from 1st sample
p
̂1 = sample proportion from 1st sample = x1/ n1

n2 independent trials from population 2

x2 = total number of successes from 2nd sample
p
̂2 = sample proportion from 2nd sample = x2/ n2

Compute the test statistic:

̂1 – p
z* = (p ̂)/[
2 p̂ (1-p̂ )(1/n1 + 1/n2)] which has an approximately standard normal
distribution and where p̂ = pooled proportion of successes= (x1+x2)/(n1+n2).
The denominator is the standard error or standard deviation of the sampling
distribution of the difference in sample proportions.

4. Calculate the p-value. The formula will depend on the form of the alternative hypothesis.
For Ha: p1 ≠ p2 [Two-tailed test] p-value = 2*P(Z > |z*| )
Ha: p1 > p2 [Upper tail test] p-value = P(Z > z*)
Ha: p1 < p2 [Lower tail test] p-value = P(Z < z*)

5. Conclusion: Make decision by comparing the p-value to .

We want to know if there is enough evidence in the data to support Ha.
If pvalue   : There is enough evidence in support of Ha, Signficant
If p-value > : There is not enough evidence to support Ha, Not significant
6. Check if the assumptions of the test are met.
1. Both samples/observations are random and from a binomial distribution
2. Two samples are from independent populations/groups
3. Both samples are large enough to have at least 5 successes and failures to be
̂1 > 5, n1(1  p
observed from each population. It suffices to check that: n 1p ̂)
1 > 5, n2p
̂2 >
5 and n2(1  p̂)
2 >5

7. Summarize the findings by writing a summary sentence.

State what your conclusion is for the particular problem regarding support or lack of support of
Ha in a sentence.

“At the  significance level, there is/is not evidence that the proportion of ______ from the
first population is different/more than/less than the proportion of ________ from the second
population [x1 = # , n1 = #, x2 = # , n2 = #, z* = #, p-value = # ]. The assumptions of the test
were/were not met since……”
 Hypothesis Testing for Two Population Means (1 and 2)

Case 1 (Section 10.2): Independent Populations and Samples

1. Define the two parameters in words:

1 = average _____________ of the first population
2 = average _____________ of the second population
Specify . If  is not given, pick a value for it. Typical values include 5%, 1% and 10%.

2. State your hypotheses.

Null Hypothesis:
Ho: 1 = 2
Alternative Hypothesis (Pick one):
Ha: 1 ≠ 2 [Two-tailed test]
Ha: 1 > 2 [Upper tail test]
Ha: 1 < 2 [Lower tail test]

3. Collect your 2 random samples and calculate the test statistic

n1 independent trials from population 1
x̅1 = sample average from 1st sample
s1 = sample standard deviation from 1st sample

n2 independent trials from population 2

x̅2 = sample average from 2nd sample
s2 = sample standard deviation from 2nd sample

Compute the test statistic:

Choose either the Two Mean Pooled T-Test or the Two Mean (Unpooled) T-Test

Two Mean Pooled T-test: Use when you assume the population standard deviations are
unknown, but equal (i.e.: Assume 1 = 2).
t* = (x̅1 – x̅2 )/Sp where the Sp = the pooled standard deviation = standard error or the
standard deviation of the difference in the sample means and
Sp = {[(n1-1) s12 + (n2-1)s22 ]/ (n1 + n2 – 2)}.
The degrees of freedom for the t distribution is df = n1 + n2 – 2.

Two Mean (Unpooled) T-test: Use when you don’t know if the population standard deviations
are equal (i.e.: Assume 1 ≠ 2).
The t*= = (x̅1 – x̅2)/S where S = [s12/ n1 + s22/n2]. And the degrees of freedom is
computed using the (complicated) Welch-Satterthwaite formula. Excel will compute this
for you.
4. Calculate the p-value. The formula will depend on the form of the alternative hypothesis
For Ha: 1 ≠ 2 [Two-tailed test] p-value = 2*P(T > |t*| )
Ha: 1 > 2 [Upper tail test] p-value = P(T > t*)
Ha: 1 < 2 [Lower tail test] p-value = P(T < t*)

5. Conclusion: Make decision by comparing the p-value to .

We want to know if there is enough evidence in the data to support Ha.
If pvalue   : There is enough evidence in support of Ha, Signficant
If p-value > : There is not enough evidence to support Ha, Not significant

6. Check if the assumptions of the test are met.

1. Two independent normal populations with unknown means 1, 2
2. Samples from each population are randomly selected
3. Distribution from each population is approximately normal.
If this is not known the test can still be used if:
n1 < 30 and n2 < 30 and histograms of each sample are approx. normal
n1  30 and n2  30 and histograms of each sample are not severely skewed and do not
have extreme outliers.

7. Summarize the findings by writing a summary sentence.

State what your conclusion is for the particular problem regarding support or lack of support of
Ha in a sentence.

“At the  significance level, there is/is not evidence that the mean ______ from the first
population is different/more than/less than the mean of ________ from the second population
[x̅1 = # , s1 = #, n1 = #, x̅2 = # , s2 = #, n2 = #, t* = #, p-value = # ]. The assumptions of the test
were/were not met since….”
Case 2 (Section 9.3): Dependent Samples (or where the two samples are
matched/paired/dependent.)

1. Define the two parameters in words:

1 = average _____________ of the first population
2 = average _____________ of the second population
Specify . If  is not given, pick a value for it. Typical values include 5%, 1% and 10%.

2. State your hypotheses.

Null Hypothesis:
Ho: 1 = 2 or diff = 0

Alternative Hypothesis (Pick one):

Ha: 1 ≠ 2 or diff ≠ 0 [Two-tailed test]
Ha: 1 > 2 or diff > 0 [Upper tail test]
Ha: 1 < 2 or diff < 0 [Lower tail test]

3. Collect your 2 samples and calculate the test statistic

Since the data is paired, you will first compute the differenced data:
x – y for each observation
Then use the 1 mean t-test using the difference data on the hypothesis test worksheet.

Then for the differenced data compute:

n = number of differences = number of paired observations
d̅ = sample average of the differences
sd = sample standard deviation of the differences

Compute the test statistic.

t* = (d̅ - 0)/(sd/n) which has an approximately a t-distribution with n-1 degrees of freedom (df)

4. Calculate the p-value. The formula will depend on the form of the alternative hypothesis
For Ha: 1 ≠ 2 or diff ≠ 0 [Two-tailed test] p-value = 2*P(T > |t*|)
Ha: 1 > 2 or diff > 0 [Upper tail test] p-value = P(T > t*)
Ha: 1 < 2 or diff < 0 [Lower tail test] p-value = P(T < t*)

5. Conclusion: Make decision by comparing the p-value to .

We want to know if there is enough evidence in the data to support Ha.
If pvalue   : There is enough evidence in support of Ha, Signficant
If p-value > : There is not enough evidence to support Ha, Not significant
6. Check if the assumptions of the test are met.
1. Two populations/samples are dependent (paired)
2. The differences have an approximately normal distribution. To check this condition,
make a histogram of sample differences and verify that the distribution is approximately
normal.
The test can still be used as long as the histogram of the differences is not severely
skewed and n is large (say, n  30).

7. Summarize the findings by writing a summary sentence.

State what your conclusion is for the particular problem regarding support or lack of support of
Ha in a sentence.

“At the  significance level, there is/is not evidence that the mean ______ from the first
population is different/more than/less than the mean of ________ from the second population
[xbar_diff = # , sDiff = #, ndiff = #, t* = #, p-value = # ]. The assumptions of the test were/were
not met.”