Testing equality of the success probabilities in two independent binomial distributions

Abstract

The aim of this thesis is to compare the power of tests of equality of the success probabilities in two independent binomial distributions for different sample sizes and for different significance levels.

The null hypothesis studied is composite. We consider four different methods of creating a p-value, the asymptotic (A) method, the estimation (E) method, the conditional (C) method and the maximization (M) method. We also consider combining the C, E, A or M method with the M method, resulting in the CM, EM, AM or M^2 p-value.