期刊名称：Sankhya. Series A, mathematical statistics and probability
印刷版ISSN：0976-836X
电子版ISSN：0976-8378
出版年度：2008
卷号：70
期号：02
页码：135--168
出版社：Indian Statistical Institute
摘要：A great upsurge in research has taken place during the last decade or so in
the area of multiple testing due to its increased relevance in modern scien-
tific investigations. A major portion of this research has been devoted to
developing methods that control error rates more meaningful as well as pow-
erful than traditional ones in the context of these investigations. The error
rates measuring false discovery proportion, particulary the false discovery
rate (FDR), and procedures controlling them have received the most atten-
tion. This article relates specifically to the FDR and is written with two
basic objectives:
(i) To put together some important developments in theory and methodology
of the FDR with finite number of hypotheses and report them with newer
formulas, proofs and insights, and
(ii) to develop an error rate extending the FDR to one that directly incor-
porates dependence among the test statistics and methods that control it.
关键词：Adpative BH methods, BH method, BY method,
false non-discovery rate, pair-wise FDR, positive dependence, multivariate
totally positive of order two, stepwise multiple testing methods
