Talk Title
High Dimensional Mediation with Interaction
Abstract
In recent years, the ever-growing attention to high-dimensional mediation inference has been prominent in diverse fields such as economics, finance, and genomic and genetic research. However, a key challenge lies in inferring the natural direct and indirect effects amidst potential interactions between the treatment and high-dimensional mediators. These interactions often lead to the well-known moderator effects, further complicated by the intricate dependence within the mediators. In this paper, we introduce a new inference procedure that addresses this challenge. By applying a non-convex penalty to the outcome model, our method efficiently identifies important mediators while handling their interactions with the treatments, which admits the guaranteed oracle property. Leveraging the oracle property, we can exploit a projection onto the mediator model, guided by the estimated important direction in the mediator space. We have established the asymptotic normality of both natural indirect and direct effects for inference. Additionally, we have developed an algorithm that utilizes the overlapped group SCAD penalty, promoting the heredity structure among the main effects and interactions, which comes with provable guarantees. Our extensive numerical studies, comparing our method with other existing approaches across various scenarios, demonstrate its effectiveness. To illustrate the practical application of our methods, we conducted a study on the impact of childhood trauma on cortisol stress reactivity. We used different DNA methylation loci as mediators, uncovering several new DNA methylation loci that remain undetected without considering interaction effects.
Bio
I am a fifth-year PhD student in the Department of Statistics at Colorado State University. Prior to joining CSU, I earned my Bachelor's degree in Economics in 2016 and my Master's degree in Applied Statistics in 2018 from Nankai University. My research interests lie in high-dimensional statistics and causal inference, and I am currently expanding my expertise into network inference. I enjoy watching soccer games in my leisure time. Passionate about advancing statistical methodologies, I am dedicated to contributing to the field through my academic pursuits and collaborative efforts.