Power Enhanced Binary Expansion Testing Procedures for Independence Detection

Abstract

In this thesis we study the Binary Expansion Testing (BET) procedure for inde- pendence detection. We first derive the theoretical optimal power achievable by the BET method when framed as a multiple hypothesis testing problem (MHTP) and showed that it brings about power enhancement over the Bonferroni procedure previ- ously used. We then create an integrated single hypothesis testing procedure aimed at reducing information loss incurred by the MHTP framing. In so doing, we derived a Weighted Alternative BET procedure that maximizes the power of the test.