SEQUENTIAL ANALYSIS OF SIMPLE HYPOTHESES WHEN USING PLAY-THE-WINNER ALLOCATION

Abstract

Play-the-winner allocations aim to place more subjects on the better treatment. The subjects enter the study sequentially. In this paper, a method for analyzing a dichotomous response, when using the randomized play-the-winner allocation and the modified play-the-winner allocation, is studied. The method is based on Wald's sequential probability ratio test. The properties of the method and of the allocation rules are studied for the case of simple hypotheses. The expected sample sizes and the expected numbers of subjects assigned to the inferior treatment are estimated by simulation.