Multiple Hypotheses Testing in Small Microarray Experiments

The approach of controlling the false discovery rate (FDR) which is less conservative than controlling the family wise error rate (FWER) has become a new and popular multiple testing procedure in microarray studies. In addition to multiple testing problems the limitation of sample size is another big challenge in this area. This thesis addresses the problem of multiple testing in the context of FDR in small samples. The first paper makes a comparison of three FDR estimation methods and the second paper investigates the effects of unequal variance between groups on the estimation of FDR in small samples using standard methods. This paper also presents a new FDR estimation procedure that deals with the above problem. Simulation results show that the new procedure works well for unequal group variance in small samples and is reliable over wider set of conditions.