Bootstrapping

Bootstrapping, a computational nonparametric technique for `re-sampling', enables researchers to draw a conclusion about the characteristics of a population strictly from the existing sample rather than by making parametric assumptions about the estimator. Using real data examples from per capita personal income to median preference differences between legislative committee members and the entire legislature, Mooney and Duval discuss how to apply bootstrapping when the underlying sampling distribution of the statistics cannot be assumed normal, as well as when the sampling distribution has no analytic solution. In addition, they show the advantages and limitations of four bootstrap confidence interval methods: normal approximation, percentile, bias-corrected percentile, and percentile-t. The authors conclude with a convenient summary of how to apply this computer-intensive methodology using various available software packages.

PART ONE: INTRODUCTIONTraditional Parametric Statistical InferenceBootstrap Statistical InferenceBootstrapping a Regression ModelTheoretical JustificationThe JackknifeMonte Carlo Evaluation of the BootstrapPART TWO: STATISTICAL INFERENCE USING THE BOOTSTRAPBias EstimationBootstrap Confidence IntervalsPART THREE: APPLICATIONS OF BOOTSTRAP CONFIDENCE INTERVALSConfidence Intervals for Statistics With Unknown Sampling DistributionsInference When Traditional Distributional Assumptions Are ViolatedPART FOUR: CONCLUSIONFuture WorkLimitations of the BootstrapConcluding Remarks