Copulas, uncertainty, and false discovery rate control

Abstract

The false discovery rate (fdr) is a powerful approach to multiple testing. However, dependence among test statistics is critical for fdr control. The way in which this dependence structure is described represents the most prominent source of uncertainty of this statistical theme. Copulas play a relevant role among the techniques used to deal with uncertainty and dependence. This paper contributes to fill an existing gap in the scientific debate by exploring the connections between the literature on fdr and that on copulas. In particular, we aim at attracting the interest of the scientific community on this topic by identifying suitable classes of nonstandard copulas which ensure that fdr control can be attained for dependent test statistics.