國立政治大學統計學系
學 術 演 講
學 術 演 講
主講人：林書勤助理教授(臺灣大學統計與數據研究所)
題 目：Causal Mediation Analysis: A Summary-Data Mendelian Randomization
Approach
時 間：民國115年3月16日 (星期一) 下午1：30
地 點：國立政治大學逸仙樓050101教室
摘 要：
Summary-data Mendelian randomization (MR) has emerged as a powerful approach for causal mediation analysis, providing an alternative to traditional methods. Two MR-based frameworks—analogous to the difference and product methods—are typically implemented using inverse-variance weighted estimation (MR-IVW). Despite their popularity, these existing approaches often suffer from limited statistical efficiency, lack of robustness, and insufficiently rigorous inference procedures.
In this talk, I will present our recent developments on improved MR-based mediation frameworks using summary-level data, commonly available from genome-wide association studies (GWAS). Specifically, we propose novel variance estimators for mediation effects, establish formal inference procedures, and introduce robust strategies to address pleiotropy. These methodological advances enhance the reliability and applicability of MR mediation analysis in practice.
In this talk, I will present our recent developments on improved MR-based mediation frameworks using summary-level data, commonly available from genome-wide association studies (GWAS). Specifically, we propose novel variance estimators for mediation effects, establish formal inference procedures, and introduce robust strategies to address pleiotropy. These methodological advances enhance the reliability and applicability of MR mediation analysis in practice.
