About the Society
Papers, Posters, Syllabi
Submit an Item
Polmeth Mailing List
Below results based on the criteria 'robust estimation'
Total number of records returned: 4
The Varying Role of Voter Information across Democratic Societies
Propensity Score Matching
Using new robust matching methods for making causal inferences from survey data, I demonstrate that there are profound differences between how voters behave in mature democracies versus how they behave in new ones. The problems of voter ignorance and inattentiveness are not as serious in mature democracies as many analysts have suggested but are of grave concern in new democracies. Citizens in mature democracies are able to accomplish something that citizens in fledgling democracies are not: inattentive and poorly informed citizens are able to vote like their better informed compatriots and hence need to pay little attention to political events such as election campaigns in order to vote as if they were attentive. The results from the U.S. (which rely on various National Election Studies) and Mexico (2000 Panel Study) are reported in detail. Results from other countries are briefly reported.
Robust Estimation and Outlier Detection for Overdispersed Multinomial Models of Count Data, with an Application to the Elian Effect in Florida
Mebane, Walter R.
overdispersed multinomial regression
We develop a robust estimation method for regression models for vectors of counts (overdispersed multinomial models). The method requires only that the model is good for most---not all---of the observed data, and it identifies outliers. A Monte Carlo sampling experiment shows that the robust method can produce consistent parameter estimates and correct statistical inferences even when ten percent of the data are generated by a significantly different process, where nonrobust maximum likelihood estimation fails. We analyze Florida county vote data from the 2000 presidential election, considering votes for five categories of presidential candidates (Buchanan, Nader, Gore, Bush and ``other''), focusing on Cuban-Americans' reactions to the Elian Gonzalez affair. We replicate results regarding Buchanan's vote in Palm Beach County. We use Census tract data within Miami-Dade County to confirm the need to take the Cuban-American population explicitly into account. The analysis illustrates how the robust method can support triangulation to verify whether a regression specification is adequate.
Detection of Multinomial Voting Irregularities
Mebane, Walter R.
generalized linear model
2000 presidential election
We develop a robust estimator for an overdispersed multinomial regression model that we use to detect vote count outliers in the 2000 presidential election. The count vector we model contains vote totals for five candidate categories: Buchanan, Bush, Gore, Nader and ``other.'' We estimate the multinomial model using county-level data from Florida. In Florida, the model produces results for Buchanan that are essentially the same as in a binomial model: Palm Beach County has the largest positive residual for Buchanan. The multinomial model shows additional large discrepancies that almost always hurt Gore or Nader and help Bush or Buchanan.
Robust Estimation of the Cox Proportional Hazards Model
Event History Modeling
Cox Proportional Hazards Model
Partial Likelihood Maximization
Iteratively-Reweighted Robust Estimation
The Cox proportional hazards model is often used with time-to-event data in political science. However, misspecification issues such as measurement error or omitted covariates can cause substantial coefficient bias when it is estimated via the conventional Partial Likelihood Maximization (PLM). Here we review an iteratively-reweighted robust (IRR) estimator of the Cox model that is proven to reduce this bias under such conditions and propose a cross-validated median fit (CVMF) test to select between PLM and IRR. Then we apply the test to data in political science. We consider several typologies of replications with respect to (1) the test's selection (PLM or IRR) and (2) the implications of IRR for the original hypotheses (less support, more support, or mixed results). Overall, we demonstrate that PLM and IRR can each be optimal, that substantive conclusions can depend on which one is used, and that the CVMF test is effective in choosing between them.