Nonparametric multiple test procedures for dose finding

Abstract

We consider identifying the minimum effective dose (MED) in a one-way layout, where the MED is defined to be the lowest dose level that is more effective than the zero-dose control. Proposed herein are two rank-based nonparametric step-down closed testing procedures that do not make order assumptions on the dose-response relationship. The corresponding p-value for the estimated MED is computed. A numerical example is given to illustrate the proposed tests. Finally, the results of a Monte Carlo study of the relative error rate and power performances are reported.