正相依条件下两步检验过程的FDR控制.docVIP

  FDR control for Two-step Procedure Under Positive Dependence# Zhuang Weiwei, Jianping Yang, Taizhong Hu, Liu Jie* 5 10 (Department of Statistics and Finance, University of Science and Technology of China, HeFei 230026) Abstract: Controlloing the familywise error-rate (FWE) is the traditional method in the multiple testing research. However, when many of tested hypotheses are rejected, controlling the FWE becomes overstrict. As the expected proportion of erroneous rejections among all rejections, the false discovery rate (FDR), introduced by Benjamini and Hochberg (1995), has been of interest in the past decade. Sarkar (2006) modified the ?idák FDR procedures under independence from a single-step to a two-step procedure. In this paper, we consider how to control the FDR for the two-step ?idák procedure, under the assumption that the test statistics are MTP2, and extend the results in Sarkar (2006). Keywords: Multiple Test; FDR; Positive dependence; MTP2 15 0 Introduction In the multiple testing problems, controlling the Type I error is a central issue. The traditional method is to control the probability of erroneously rejecting even one of the true null hypotheses, the familywise error-rate (FWE). However, when many of tested hypotheses are rejected, 20 25 30 implying that many hypotheses are not true, controlling the FWE becomes overstrict, and controlling the proportion of errors seems appropriate. The false discovery rate (FDR), as the expected proportion of erroneous rejections among all rejections, was introduced by Benjamini and Hochberg (1995)[1]. In the past decade, a great number of articles dealted with controlling the FDR have been contributed; see Benjamini and Liu (1999)[2], Benjamini and Yekutieli (2001) [3] , Sarkar (2002, 2004, 2006) [5] [6][7] , Blanchard and Roquain (2009) [4], etc. Consider simultaneous testing of n null hypotheses H1,K,H n . Let Q denote the proportion of false positive (Type I errors) among the rejected null hypotheses