Assignment #9
- PPol 604
- Due: Thursday, 21 March 2013
Type up your answers. Give proper credit to those you work with and/or the text(s).
Solve the following problems. Show all of your work, but keep your answers concise.
Highlight your (final) answer
to distinguish it from your other numbers and text. Include a copy of your input
(e.g. do file) or output (e.g. log file),
when it is an appropriate way to show your work.
However, do not include unnecessary output (i.e. no data dumps), and format any output
so that it is easily readable.
An appropriate time to include output is when you put your results
in a table--if your results are wrong, then the grader has no idea how you came to your
conclusions (i.e. give partial credit) unless you provide some output. Explanation
includes statistical and substantive explanation (explain so that a statistical
layperson can understand it, and so that a
statistical analyst will see your erudition).
- {50} [from Zorn and Van Winkle 2000 via Singer and Willett 2003] Singer and Willett present data
on the time to retirement or death for U.S. Supreme Court justices. The
data set found
here contains data on the time to retirement or death for U.S. Supreme Court justices. for this problem,
retirement constitutes a failure, and death constitutes censoring.
Reproduce the data analysis of CGGM, Chapters 8-9, using this data set, including:
- Kaplan-Meier method of estimating survivor function (including graph)
- Using age (of first appointment) as a covariate, plot within-group K-M survivor functions
for high age and low age (greater and lower than the median, respectively)
- Nelson-Aalen method of estimating cumulative hazard function (including graph)
- Using age (of first appointment) as a covariate, plot within-group N-A cumulative hazard functions
for high age and low age (greater and lower than the median, respectively)
- Kernel-smoothed estimates of hazard function, using different bandwidths (bwidth(#) option)
- Using age (of first appointment) as a covariate, plot within-group smoothed hazard functions
for high age and low age (greater and lower than the median, respectively)
- Estimate the median survival time for high and low age
- Test for the equality of survivor functions for high and low age
- Estimate a Cox model, using age as a binary covariate (as above) and the Efron method for ties; interpret the hazard ratio
- Estimate a Cox model, using age as a continuous covariate (Efron method); interpret the hazard ratio
- Graph survivor, cumulative hazard, and hazard functions for two different ages from the Cox model (using continuous age)
- {50} [from King, Alt, Burns, and Laver 1990 via Box-Steffensmeier and Jones 2004]
Why do different cabinets fail more quickly than others? The data set found
here contains data on
the survival of 313 (mostly) European cabinet governments if 15 countries.
DURAT
is how long the cabinet lasts (in months). We will first replicate King et al.'s analysis with seven covariates. The first, INVEST, is a binary variable denoting whether
or not an initial confirmatory vote is required by the legislature. This
legal requirement is a "hurdle" that governments must
overcome. The second covariate, POLAR, measures the percentage
of support for extremist parties. The idea here is that as the support for
extremist parties increases, the degree of polarization in the government will
increase. The third covariate of interest
is a binary indicator denoting whether or not the government has a numerical
majority: NUMST2. The fourth covariate is a count of the number of attempts to form a
government prior to the official formation of the government: FORMAT. The
fifth covariate is a binary indicator denoting if the government was formed immediately
after an election: ELTIME2 (as opposed to forming in the middle of an election cycle, which means
a cabinet failed but the new cabinet formed without an election). The sixth
covariate is a binary indicator denoting whether or not the government is serving
as a "caretaker" government: CARETK2. And the seventh covariate is a measure of fractionalization: FRACT.
The variable, CIEP12, denotes whether a cabinet
failed. It is coded 1 if the cabinet failed, and 0 if the cabinet did not fail, or ended within 12 months of the
"constitutional interelection period," which requires that an election be called
and the cabinet disbanded.
- Use tools learned in the previous problem to analyze the data. Use the Efron method for ties. Interpret the results of your final model. (You do not need to write a professional report.)
- Going beyond the tools of the previous problem, assess whether accounting for dependent observations affects the analysis:
- Cluster the standard errors by country. What is the rationale for this? Comment on any changes from your final model.
- Include a fixed effect for each country. What is the rationale for this? Comment on any changes from your final model. Do the fixed effects improve model fit (using AIC/BIC)?
- Include a random effect (shared frailty) for each country. What is the rationale for this? Comment on any changes from your final model. Do the random effects improve model fit (using AIC/BIC)? Which country's cabinets are most likely to fail (controlling for included covariates)? Which is least likely?
- Which approach(es) would you choose? Why?
- A theory in the political science literature states that minority cabinet governments are qualitatively different than majority governments. Assess this theory going beyond the tools of the previous problem:
- First, using your final model, how are minority governments different than majority governments? What is the rationale for this approach? Compare the survivor/hazard/cumulative hazard for minority and majority governments.
- Run separate Cox models for minority and majority governments. What is the rationale for this? Compare the survivor/cumulative hazard and the other covariates.
- Stratify the Cox model between minority and majority governments. What is the rationale for this? Compare the survivor/cumulative hazard.
- Which approach would you choose? Why?
Back to
Assignments
page