Dealing With the Multiple Testing Problem in Race and Sentencing Research: The Value of the Bootstrap Correction Method

Marc L. Swatt, University of Nebraska at Omaha
Miriam A. DeLone, University of Nebraska - Omaha

The purpose of this paper is to examine the multiple testing problem in the race and sentencing literature. Specifically, this paper will address the use of interaction terms (both additive interactions and split-data methods) and how they may potentially alter interpretations of statistical significance in regression models. Resampling procedures will be described as an alternative to single step adjustments to adjusting for the multiple testing problem. The data that will be used come from the Federal Sentencing Guidelnes Dataset, specifically all cases of robbery will be included in this analysis. The initial dataset will be treated as a population and analyzed. Then a random sample of 5% of the population will be drawn and analyzed. Finally, the bootstrap resampling procedure will be used to calculate regression coefficients and adjusted p-values. The results of these different analyses will be compared to the results found in the population. Implications for research on race and sentencing will be discussed.

(Return to Program Resources)

Updated 05/20/2006