Primary Submission Category: Design-Based Causal Inference
Causal subgroup discovery with error control
Authors: Yao Zhang, Zijun Gao,
Presenting Author: Yao Zhang*
In randomized controlled trials, researchers are often interested in understanding the heterogeneity of treatment effects through interpretable subgroup analysis. In this paper, we propose a new subgroup analysis method that leverages regression models to discover patient subgroups with significant treatment effects. This method then validates these subgroups using the randomness of treatment assignment, thereby limiting the number of false discoveries. Through experiments, we demonstrate that this new approach not only identifies promising subgroups effectively but also provides more reliable insights for developing personalized treatment plans for patients with varying conditions.