Sequential Multiple Testing for Variable Selection to Control Error Rate

• Activity Number: 543
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Hailu Chen* and Cui Xinping
• Companies: UC Riverside and UC Riverside
• Keywords: Multiple testing ; FWER ; FDR ; LASSO

Abstract

Lockhart et al. (2014) proposed a simple covariance test for testing the significance of the predictor variable that enters the current lasso model along the lasso solution path. In this paper, we propose a hybrid sequential multiple testing procedure using covariance test p-values, which has a good power properties with error rate controlled at desired level. Specially, we consider the full underlying hypotheses and the error rate control within each step as well as across all steps along the LASSO solution path. To control FWER at desired level, we propose hybrid-Bonferroni, Hochberg and Sime's methods and compare with Single hypothesis. To control FDR, we propose hybrid-BH and compare it with Single hypothesis BH method and StrongStop algorithm. Simulation studies show that our proposed procedures have higher power with both FWER and FDR controlled at desired level. We also apply the proposed procedure for genetic real data analysis to evaluate our method and compare with other methods discussed above.