Searching for subgroup-specific associations while controlling the false discovery rate

TL;DR

This paper presents a new method for subgroup-specific association discovery in high-dimensional data, controlling false discoveries and leveraging machine learning models without sample splitting.

Contribution

It extends the model-X knockoff filter to enable automated, subgroup-aware hypothesis testing with rigorous FDR control, improving interpretability and efficiency.

Findings

Effective in simulations and real data experiments

Controls false discovery rate in subgroup analyses

Leverages machine learning models for hypothesis generation

Abstract

This paper introduces an innovative method for conducting conditional independence testing in high-dimensional data, facilitating the automated discovery of significant associations within distinct subgroups of a population, all while controlling the false discovery rate. This is achieved by expanding upon the model-X knockoff filter to provide more informative inferences. Our enhanced inferences can help explain sample heterogeneity and uncover interactions, making better use of the capabilities offered by modern machine learning models. Specifically, our method is able to leverage any model for the identification of data-driven hypotheses pertaining to interesting population subgroups. Then, it rigorously test these hypotheses without succumbing to selection bias. Importantly, our approach is efficient and does not require sample splitting. We demonstrate the effectiveness of our method through simulations and numerical experiments, using data derived from a randomized experiment featuring multiple treatment variables.