Selecting the better binomial population with unequal sample sizes

Abstract

Ihe problem of selecting the better binomial population is considered in the case of unequal sample sizes. A class of decision rules, which depend on only two parameters regardless of sample sizes, is formulated. Extensive numerical studies indicate that the minimax decision rule within the proposed class possesses many properties consistent with those of a globally minimax decision rule. In particular, the restricted minimax decision rule is demonstrated to be globally minimax in the limit as one sample size tends to infinity as well as in certain small sample configurations.