Speaker：Shan Luo (Shanghai Jiao Tong University)

Time：2023-5-26, 14:30

Location：Conference Room 109 at Experiment Building at Haiyun Campus

Abstract:

Causal mediation analysis is a fundamental statistical approach to understanding the effect of exposure on an outcome. In this paper, we investigate group inference for high-dimensional mediation models by considering the mediators in an interested group jointly or individually. For both situations, we construct suitable test statistics and establish their asymptotic distributions. A simple estimator for the joint group indirect effect is also introduced. Its asymptotic normality is also established. Extensive numerical studies demonstrate that our proposed methods outperform recent representative approaches. We also apply our methods to analyse how DNA methylation operates in the regulation of human stress reactivity impacted by childhood trauma.