时间：12月16日（周三）上午10:30-11:30
地点：统计楼105报告厅
报告人：Tony Cai Department of Statistics The Wharton School University of Pennsylvania
主题：Large-Scale Two-Sample Inference for Sparse Means
简介：The conventional approach to 2-sample multiple testing is to first reduce the data matrix to a single vector of p-values and then choose a cutoff along the rankings to adjust for multiplicity. However, this inference framework often leads to suboptimal multiple testing procedures due to the loss of information in the data reduction step. In this talk, we introduce a new method to large-scale multiple testing for two sparse means. The problem is studied in a decision-theoretic framework and both oracle and data-driven procedures for FDR control are developed. The proposed oracle procedure employs a covariate-assisted ranking and screening (CARS) technique, and is shown to be optimal. A data-driven procedure is then developed to mimic the oracle procedure and its asymptotic properties are established.