Online rules for control of false discovery rate and false discovery exceedance (Q1750278): Difference between revisions

For a set of null hypotheses in large scale hypothesis testing problems, consider the false discovery rate (FDR) defined by the expected proportion of discoveries that are false positives. Here, methods for online control of false discovery rate are studied: Let \(H_1\), \(H_2\), \(H_3,\dots\) denote a sequence of null hypotheses with input \(p\)-values \(p_1\), \(p_2\), \(p_3,\dots\). A test procedure is suggested involving at each step the decision -- depending on the previous actions -- whether to reject/accept the current hypothesis using test significance levels \(\alpha_1\), \(\alpha_2\), \(\alpha_3,\dots\). The aim is then to determine test procedures having a false discovery rate below a given level \(\alpha\). Examples and numerical simulations are given.