Controlling the false discovery rate (FDR) is a powerful approach to multiple testing, with procedures developed with applications in many areas. Dependence among the test statistics is a common problem, and many attempts have been made to extend the procedures. In this paper, we show that a certain degree of dependence is allowed among the test statistics, when the number of tests is large, with no need for any correction. We then suggest a way to conservatively estimate the proportion of false nulls, both under dependence and independence, and discuss the advantages of using such estimators when controlling the FDR. © 2007 Board of the Foundation of the Scandinavian Journal of Statistics.
Some results on the control of the false discovery rate under dependence / Farcomeni, Alessio. - In: SCANDINAVIAN JOURNAL OF STATISTICS. - ISSN 0303-6898. - 34:2(2007), pp. 275-297. [10.1111/j.1467-9469.2006.00530.x]
Some results on the control of the false discovery rate under dependence
FARCOMENI, Alessio
2007
Abstract
Controlling the false discovery rate (FDR) is a powerful approach to multiple testing, with procedures developed with applications in many areas. Dependence among the test statistics is a common problem, and many attempts have been made to extend the procedures. In this paper, we show that a certain degree of dependence is allowed among the test statistics, when the number of tests is large, with no need for any correction. We then suggest a way to conservatively estimate the proportion of false nulls, both under dependence and independence, and discuss the advantages of using such estimators when controlling the FDR. © 2007 Board of the Foundation of the Scandinavian Journal of Statistics.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.