ZAP:<i>Z</i>-Value Adaptive Procedures for False Discovery Rate Control with Side Information

Abstract Adaptive multiple testing with covariates is an important research direction that has gained major attention in recent years. It been widely recognised leveraging side information provided by auxiliary can improve the power of false discovery rate (FDR) procedures. Currently, most such procedures are devised p-values as their main statistics. However, for two-sided hypotheses, usual data processing step transforms primary statistics, known p-values, into not only leads to a loss carried but also undermine ability assist FDR inference. We develop p-value based covariate-adaptive (ZAP) methodology operates on intact structural encoded jointly and covariates. seeks emulate oracle procedure via working model, its rejection regions significantly depart from those adaptive approaches. The key strength ZAP control guaranteed minimal assumptions, even when model misspecified. demonstrate state-of-the-art performance using both simulated real data, which shows efficiency gain be substantial comparison p-value-based methods. Our implemented R package zap.