Computational improvements in bootstrap ranking & selection procedures via multiple comparison with the best

General-purpose ranking and selection (R&S) procedures using bootstrapping were investigated by Lee and Nelson in WSC ´14; their work provides the seminal idea for this study. Here we present bootstrap R&S procedures that achieve significant computational savings by exploiting multiple comparison with the best inference. We establish the asymptotic probability of correct selection for the new procedures, and report some experiment results to illustrate small-sample performance, both in attained probability of correct selection and computational efficiency relative to the procedures in Lee and Nelson.