PUB AF 8180: Research Methods and Inquiry in Public Affairs

Harry S Truman School of Government and Public Affairs

University of Missouri-Columbia

Practice for Quiz 2

1. During a recent conversation with her grandfather, a student enrolled in an MPA

program hears the common line of “you young kids today just don’t work as hard as my
generation did.” Wondering if her grandfather’s statement has any merit, she decides to
investigate. She finds that according to the Federal Reserve Bank, the average hours
worked per week per worker in 1950 was 42.4. She takes a sample of 50 workers and
finds the mean hours worked of 42.1 with a standard deviation of 5.9.

a. What are the null and alternative hypotheses?

The null hypothesis is that the hours worked in the sample is equal to 42.4, our known
parameter.

The alternative hypothesis is that the hours worked in the sample is less than 42.4. This is
because our student’s grandfather things young people today “don’t work as hard.”

b. Calculate the appropriate test statistic.

This is a single sample t-test. Our first step is to calculate the standard error for the sample.

s.e. = s / √n = 5.9 / √50 = 0.83

Then, we calculate the test statistic

t = (𝑥𝑥̅ -μ)/s.e. = (42.1-42.4)/0.83 = -0.36

c. Based on your calculations, what conclusion do you reach?

We compare the test statistic to a critical value of -1.96. Because our test statistic is not lower
than this critical value, we fail to reject our null hypothesis.

Our student should tell her grandfather that kids today are working just as many hours as “back in the
day.”
 2. A teacher is interested in whether the use of a study guide will improve test scores. The
table below shows scores on a vocabulary test before using a new study guide and
tests on a similar vocabulary test after using the study guide for a random sample of 5
students.

Person # Before Study After Study

Guide Guide
1 88 89
2 92 90
3 85 87
4 80 84
5 83 84

a. Develop a null and alternative hypothesis for the teacher’s research question.

Our null hypothesis is that scores before the study guide are equal to scores after the study
guide, or that the difference = 0

Our alternative hypothesis is that scores increased after the study guide, or that before is less
than after (or that before – after < 0).

b. Calculate the appropriate test statistic.

We have a paired t-test here. We want to keep the observations matched. Our process
is to calculate the difference for each student, then calculate the mean, standard
deviation, and standard error for the difference.

The mean difference is -1.2; the standard deviation of the differences is 2.17, and the
standard error of the difference is 2.17. I did this in excel, so you can see the solution
table below. I promise, there is not a paired t-test calculation question on the quiz.
 Before After
Differen (x-
Person # Study Study x-xbar
ce xbar)^2
Guide Guide
1 88 89 -1 0.2 0.04
2 92 90 2 3.2 10.24
3 85 87 -2 -0.8 0.64
4 80 84 -4 -2.8 7.84
5 83 84 -1 0.2 0.04
Mean--> -1.2 18.8 <-- sum
2.17 <-- st. dev
0.97 <-- s.e.

Our test statistic is calculated as (-1.2-0)/0.97 = -1.24

c. Based on your calculations, what conclusion do you reach about the effectiveness of
the study guide?

Because our sample size is small, we need to use the t-distribution table to find the critical
value. The critical value associated with 95% confidence and 4 degrees of freedom is +/-
2.78.

Because our test statistic is not less than the critical value of -2.78, we fail to reject the null
hypothesis. Our data does not show any evidence that the study guide increased tests
scores.
 3. A professor would like to compare scores on a test between two sections of her class.
The first section has 75 students and the second has 200 students. The mean score and
standard deviation are presented in the table below. The professor wants to know if
there is any significant difference in scores between the two groups.

Section 1 Section 2
Sample size 75 200
Mean score 74.5 70.6
Standard deviation 31.4 26.3

a. What are the null and alternative hypotheses?

The null hypothesis is that the score for section 1 = the score for section 2.
The alternative hypothesis is that the scores for the two sections are not equal (we have no
basis to predict one section being larger than the other)

b. Calculate the appropriate test statistic.

This is a two-sample ttest with independent samples. First, we need to calculate a standard
error for each of our samples, and then an overall standard error.

se(1) = 31.4/√75 = 3.64

se(2) = 26.3/√200=1.86

se(overall) = √(se1^2 = se2^2)  4.09

Our test statistic is calculated as (74.5-70.6)/4.09 = 0.95

c. Based on your calculations, what conclusion do you reach?

With a large sample (df = 75+200-2=273), we use the critical values of +/- 1.96

Because our test statistic is not greater than the critical value of 1.96, we fail to reject our null
hypothesis. There is not evidence with these samples that the two sections are performing
differently on the test.
 4. Using the National Election Studies dataset, I’m interested in examining how men and
women feel about certain issues. I suspect that men feel more positively about labor
unions since the rate of union membership is slightly higher for males. I run the following
test in Stata.

Does this output support my hypothesis? Why or why not?

First, with this question, we need to carefully craft our null and alternative hypothesis and be sure
we examine the correct alternative hypothesis output. So, yes, this is kind of a tricky question!

Our null hypothesis is that the thermometer for men = thermometer for women

Our alternative hypothesis is that the thermometer for men > thermometer for women, or that
thermometer(men) – thermometer(women) > 0

Therefore, we have a one sided test, and we look at the ha: diff > 0

Since that p value is not less than 0.05, we fail to reject the hull hypothesis.

But wait! Look at those other p values, the ones for the other possible alternative hypotheses….So,
not only is my hypothesis not supported, the data actually show that women have significantly
more positive feelings about unions than men.
 5. Using the same dataset from #4, I run a cross tabulation between respondents’ political
ideology and their interest in campaigns.

a. What percent of conservatives have a high interest in campaigns? Look at the row total
here  548/1131=48.45
b. What percent of those with high campaign interest are moderate? Look at the column
totals here  93/1083=8.59
c. What is the null and alternative hypothesis for the chi square test?

The null hypothesis for the chi square test is that there is no relationship between political ideology
and interest in political campaigns

The alternative hypothesis for the chi square test is that there is a statistical relationship between
political ideology and interest in political campaigns.

d. Based on the chi square test, does there appear to be a statistical relationship between
these two variables?
Thankfully, you don’t have to calculate a chi square test statistic (you’re welcome 😊😊) Stata does
that for us – the statistic is 21.1740 and the p value is 0.000.

There are two ways we can evaluate the statistical significance. First, we can compare the test
statistic to a critical value from a chi square table. With 4 degrees of freedom and 95 %
confidence, the critical value is 9.49. Since our chi square statistic is greater than that critical
value, we reject the null hypothesis.

A more straightforward way is to evaluate the p value. Since it is lower than 0.05, we can reject the
null hypothesis.

So, we conclude from the chi square test that there is a significant relationship between political
ideology and interest in political campaigns. That’s all we can say with the chi square test, for
further analysis, we could use a measure of association (gamma) to measure the direction and
strength of the relationship.