49
18

False Discovery Rate Control under Archimedean Copula

Abstract

We are considered with the false discovery rate (FDR) of the linear step-up test φLSU\varphi^{LSU} considered by Benjamini and Hochberg (1995). It is well known that φLSU\varphi^{LSU} controls the FDR at level m0q/mm_0 q / m if the joint distribution of pp-values is multivariate totally positive of order 2. In this, mm denotes the total number of hypotheses, m0m_0 the number of true null hypotheses, and qq the nominal FDR level. Under the assumption of an Archimedean pp-value copula with completely monotone generator, we derive a sharper upper bound for the FDR of φLSU\varphi^{LSU} as well as a non-trivial lower bound. Application of the sharper upper bound to parametric subclasses of Archimedean pp-value copulae allows us to increase the power of φLSU\varphi^{LSU} by pre-estimating the copula parameter and adjusting qq. Based on the lower bound, a sufficient condition is obtained under which the FDR of φLSU\varphi^{LSU} is exactly equal to m0q/mm_0 q / m, as in the case of stochastically independent pp-values. Finally, we deal with high-dimensional multiple test problems with exchangeable test statistics by drawing a connection between infinite sequences of exchangeable pp-values and Archimedean copulae with completely monotone generators. Our theoretical results are applied to important copula families, including Clayton copulae and Gumbel copulae.

View on arXiv
Comments on this paper