BOOTSTRAP STATISTICAL-INFERENCE - EXAMPLES AND EVALUATIONS FOR POLITICAL-SCIENCE

Theory: Bootstrapping is a nonparametric approach to statistical infer ence that relies on large amounts of computation rather than mathemati cal analysis and distributional assumptions of traditional parametric inference. It has been shown to provide asymptotically accurate infere nces for a wide variety of statistics. Hypothesis: Bootstrapping may m ake more accurate inferences than the parametric approach under two ge neral circumstances: 1) when the assumptions of parametric inference a re not tenable, and 2) when no parametric alternative exists for a pro blem. Methods: Monte Carlo simulation is used to test the performance of bootstrap and parametric confidence intervals for both types of sit uations in which bootstrapping is hypothesized to be superior to param etric inference. A single data example is used to illustrate the use o f the bootstrap: a seats/votes model of U.S. House elections from 1932 to 1988. Results: My central conclusions are that in the cases examin ed: 1) bootstrap confidence intervals are at least as good as the para metric confidence interval and sometimes better, 2) OLS parametric con fidence intervals do not perform too badly when the model error is non -normal, especially as sample size increases, and 3) when no parametri c alternative exists, the bootstrap provides a reasonable method of ma king statistical inferences.