Empirical Power of Procedures for Comparing two Groups on p Variables

Abstract

Monte Carlo results are used to evaluate several procedures for discriminating between two groups on two or more variables. Depending on the conditions examined, either a multiple testing version of Hotelling’sT ² or a Bonferroni procedure is shown to be the most powerful for detecting at least one true difference between the groups. A multiple Bonferroni procedure is shown to be generally superior in power for detecting all true differences.