Robust selection procedures based on vector ranks (Q1072286)

This article examines the situation wherein data arise from a 2-way analysis of variance type setting, while most of the selection literature assumes a 1-way setting. Thus, such settings, which are known to be important in practice, as those where there are block effects, are included. The problem of interest is the selection of the t best populations. A vector within blocks procedure PV is compared to the normal theory procedure MP. PV addresses an important problem, which (in the case of ordinal data) work on robust selection by a number of authors does not apply to. It is hoped that the rigorous treatment given will stimulate further work in this area.