Abstract
In this article the problem of comparing two independent binomial populations is considered. It is shown that the test based on the confidence interval p value of Berger and Boos often is uniformly more powerful than the standard unconditional test. This test also requires less computational time.