Generalized stepwise procedures for controlling the false discovery rate
Open Access
Author:
Roths, Scott
Graduate Program:
Statistics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
June 22, 2011
Committee Members:
G Jogesh Babu, Dissertation Advisor/Co-Advisor G Jogesh Babu, Committee Chair/Co-Chair Debashis Ghosh, Committee Chair/Co-Chair Bing Li, Committee Member Stephen George Simpson, Committee Member
Keywords:
false discovery rate stepwise procedures
Abstract:
Among the most popular procedures used to control the false discovery rate (FDR) in large-scale multiple testing are the stepwise ones, where the marginal p-values are ordered and compared with specified cutoffs, according to a stopping rule. Starting with the most significant p-value, the stepdown procedure rejects each null hypothesis as long as the corresponding cutoff is not exceeded. The stepup procedure starts with the least significant p-value and proceeds in the opposite direction and accepts each null hypothesis, provided the p-value does not exceed its cutoff. These procedures have been shown to control the FDR under certain types of dependencies, including the kind relevant to multiple comparisons with a control. A generalization of these stepwise procedures allows the stepdown procedure to continue rejecting as long as the fraction of p-values not exceeding their cutoffs is sufficiently large is proposed. The stepup procedure is similarly generalized. For appropriate choices of this fraction bound, increased power may be obtained. The proposed method is illustrated with simulated and real data.
