Part 1: Emergency Room Visits and the Full Moon
Movies about werewolves are based on the idea that behavior changes when there's a full moon. This is where the term "lunacy" comes from. Studies investigate whether it might have something to do with the moon's impact on water on Earth. Maybe the moon creates physiological and psychological effects on some people. Consequently, there might be more "psychiatric hospital admissions, suicides, homicides, emergency room calls, traffic accidents" etc.
A group of researchers tests the hypothesis that emergency room visits increase when there's a full moon. The dependent variable is "cases" which captures the number of emergency room cases per night. The independent variable is "full moon" coded as 1 when there is a full moon and zero otherwise. Other variables in the dataset are as follows:
Table 1Variable Definitions
|
Holiday
|
1= if it is a holiday, 0 otherwise.
|
|
Friday
|
1= if it is a Friday, 0 otherwise.
|
|
Saturday
|
1= if it's a Saturday, 0 otherwise.
|
The researchers obtain the results in Table 2.
Table 2The Effect of the Full Moon on Emergency Room Visits
|
Independent Variable
|
|
|
Fullmoon
|
1.743
(3.980)
|
|
|
|
|
Constant
|
100.5068
|
|
R2
|
0.0007
|
|
N
|
229
|
The dependent variable is Cases, the number of emergency room visits.
Questions:
1. Write out the SRF for the results in Table 2.
2. Is the coefficient on Fullmoon statistically significant? Explain how you arrived at your answer.
3. Interpret the constant in Table 2 and also explain if it has a meaningful interpretation.
4. Interpret the R2.
5. Can the researchers confidently conclude that on days when there is a full moon emergency rooms will experience a higher number of visits caused by the erratic behavior the full moon induces? Explain why or why not.
6. Using the dataset described in Part 1, how might you improve the model presented in Table 2.
Part 2: Congressional Elections and Voter Turnout
It is well established that voter turnout in U.S. congressional elections is lower than in presidential elections. Using data on the 1978 congressional elections, political scientist, Franklin D. Gilliam, Jr. sought to study the reasons for this. The hypothesis is that when campaigns spend more on an election and races are competitive turnout is higher.
The dependent variable is "turnout" defined as "the ratio of the two-party vote in the 1978 congressional election to the district's voting age population." The independent variables are described in Table 3. Table 4 shows an edited version of Gilliam's findings. Using the information from Tables 3 and 4, answer the questions below.
Table 3: Variable Definitions
|
Marginality
|
Dummy variable coded 0/1 capturing voters' perceptions of competition, where 0= perfect competition, 1=otherwise, it is the"difference between the Democratic and Republican votes, divided by the total two-party vote, in absolute values."
|
|
Party Competition
|
Dummy variable coded 0/1, where 0= perfect competition and 1= lack of competition "the ratio of the sum of differences in the two- party vote in two previous elections between 1972 and 1976 to the sum of the total two-party vote over the same period, in absolute values."
|
|
Education
|
"Is the proportion of the voting-age population with some college education."
|
|
Campaign Spending
|
"Is the ratio of total two-party expenditures (in cents) to eligible voters."
|
|
Region
|
"Is a dummy variable, with a value of one for districts located in the states of the "solid South" and a value of zero elsewhere."
|
|
Race
|
"The ratio of blacks over 18 to the voting-age population;
|
|
Urbanism
|
is "the proportion of the district's population residing in an urban area."
|
Table 4: Impact of Spending and Competition on Congressional Election Turnout
7. What is the t-statistic on Campaign Spending? Does it indicate the variable is significant at the 5% and 1% levels?(Table 4)
8. Interpret the coefficient on Campaign Spending. Discuss its magnitude, direction, statistical and substantive significance.
9. Do the findings suggest that turnout is higher among individuals with a college education? Explain.
10. What could you conclude about Urbanism and turnout?
11. Are states from the South more or less likely to experience high turnout in congressional elections?
12. Using the findings in Table 4, can we confidently conclude that this model explains why turnout in congressional elections is low? Explain why or why not.
Part 3: Smoking and Pregnancy
Does smoking during pregnancy affect the baby's health? Researchers want to know if smoking impacts a baby's health in a negative fashion. To measure infant health they use birth weight (bweight), because "a birth weight that is too low can put an infant at risk for contracting various illnesses." The dependent variable is bweight, and the independent variables are listed below in Table 5.
The researchers run various models and obtain the results in Table 6.
Table 5 Variable Definitions
|
Variable
|
Description
|
|
bweight
|
Birth weight in pounds
|
|
faminc
|
1988 family income, $1000s
|
|
fatheduc
|
father's years of education
|
|
motheduc
|
Mother's years of education
|
|
cigs
|
Cigarettes smoked per day while pregnant
|
|
parity
|
Birth order of child
|
|
cigtax
|
Cigarette tax in home state, 1988
|
|
moth_college
|
=1 if mother has 16 year of education or more
|
|
fath_college
|
=1 if father has 16 year of education or more
|
Table 6 The Effect of Smoking on Birth Weight
|
|
Model 1
|
Model 2
|
Model 3
|
Model 4
|
|
Dependent
Variable:
|
Birth
Weight
|
Birth
Weight
|
Birth
Weight
|
Birth
Weight
|
|
Cigs
|
-.032
( .007 )
|
-.028
(.007)
|
-.038
( .007)
|
-.039
(.008)
|
|
Faminc
|
-
|
.005
(.002)
|
.005
(.002)
|
.005
(.004)
|
|
Motheduc
|
-
|
-
|
-.26
(.109)
|
-.28
(.109)
|
|
Fatheduc
|
-
|
-
|
.12
(.106)
|
.12
(.19)
|
|
Cigtax
|
-
|
-
|
.007
(.005)
|
.007
(.005)
|
|
Parity
|
-
|
-
|
-
|
.084
(.048)
|
|
Constant
|
7.470676
|
7.285195
|
7.243484
|
7.102938
|
|
|
|
R2
|
0.0223
|
0.0299
|
0.0405
|
0.0438
|
|
n
|
999
|
999
|
999
|
999
|
Notes: Infant health is measured with the variable bweight.
13. Using Model 1, test the hypothesis that the coefficient on cigs is zero against the alternative that it is non-zero at the 5% significance level.
14. Using Model 1, interpret the magnitude of the coefficient, cigs, in real-world terms.
15. Interpret the constant in Model 1.
16. Compare Model 1 to Model 2, is famincan omitted variable? Explain.
17. Using Model 3, what is the expected weight of a newborn whose mother smoked 15 cigarettes a day, had no college education, and lived in a household with a family income of $47,000? The father did not have a college education and there is no state tax on cigarettes.
18. Does the R2 in Model 4 provide information on whether the coefficients are biased?
19. Which model is best and why?
20. Can we confidently conclude that the model you selected in question #19 fully explains infant health as measured by birth weight?
Part 4: Religiosity and Support of Marriage Equality
Researchers were interested in the extent to which religiosity influenced attitudes towards marriage equality in 2004. Using data from the National Election Pool General Election Exit Polls from 2004, they tested the impact of religiosity on attitudes on attitudes towards marriage equality. They tested the following hypotheses:
- Religious individuals are more likely to identify with the Republican Party since the party often employs religion in platforms and speeches.
- Individuals who identify as Republican are more likely to oppose gay marriage because the party famously opposes the reform in 2004.
- Highly religious individuals are more likely than others to oppose gay marriage because they find it inconsistent with religious doctrine.
The dependent variable is Gaymarriage, a binary variable where 1 indicates opposition towards marriage, and 0 otherwise . The independent variables are as follows:
Table 7 Variable Definitions
|
Independent Variables
|
Definition
|
|
Gaymarriage
|
= 1 if respondent opposes gay marriage, 0 otherwise.
|
|
Religious
|
= scale of 0 to 4, where increasing values indicates increased adherence to religion.
|
|
Partyid
|
= 1 if respondent is Republican, 0 otherwise.
|
|
Interact
|
Interaction term, religious*partyid
|
Consider Models 1 and Model 2 in the Stata outputs below to answer the questions: (questions #21-25 pertain to Part 4).
21. Using Model 2, write out the SRF among non-Republicans.
22. Using Model 2, what is the effect of a one unit increase in religious on gaymarriage?
23. Do the results in Models 1 and 2 suggest that non-Republicans are more likely to oppose marriage equality? Explain.
24. Explain the change in the coefficient on partyidfrom Model 1 to Model 2. What might you conclude from this?
25. Does Model 2 indicate an interactive effect between religious and partyidon gaymarriage? Explain.
Extra credit: (2.5 points) If you get a perfect score on the exam, the extra credit is applied towards the total grade points.
Using any of the datasets used in this course (excluding the final exam), state an original hypothesis you would like to test.
- Define the dependent and independent variables (needs to be more than one independent variable).
- Write out the SRF.
- Explain the methodology you would use.
- Discuss any limitations to your approach, pay attention to the data and methods.
- Imagine that you obtained findings. Include a sentence or two translating these for broad audiences.
Attachment:- Exam.rar