报告题目：普遍未知混杂下高维中介分析的双重纠偏估计与推断

报告人：秦国友 教授（复旦大学）

报告时间：2025年10月30日（周四）上午10:00

报告地点：数学统计学院145学术报告厅

报告摘要：Mediation analysis is a powerful tool for elucidating the causal mechanisms by which exposures influence outcomes through mediators. However, conventional approaches often yield biased estimates of mediation effects in settings involving high-dimensional exposures and mediators, as well as pervasive hidden confounders. To address these challenges, we propose a deconfounded-debiased estimation for high-dimensional mediation analysis to estimate direct and indirect effects based on the difference-in-coefficients strategy. This approach effectively corrects biases arising from both pervasive hidden confounders and high-dimensionality. We establish the asymptotic normality of the proposed estimators for both direct and indirect effects and develop hypothesis testing procedures that achieve asymptotically controlled type I error probability and full power. Simulation experiments demonstrate the method’s superior finite-sample performance in estimating mediation effects, especially in the presence of pervasive hidden confounding. We also applied the proposed method to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database to explore the association between serum metabolites and Alzheimer’s disease cognitive severity mediated by DNA methylation levels.

报告人简介： 秦国友，博士，复旦大学教授、博士生导师，公共卫生学院生物统计学教研室主任。主要从事生物统计学方法学和应用研究，包括真实世界研究和因果推断，临床试验、针对复杂数据、复杂统计模型的统计方法创新，以及生物统计学方法在医学和公共卫生领域的应用。纵向数据分析相关研究工作获得了2014年教育部高等学校科学研究优秀成果二等奖。在BMJ、Plos Medicine、JAMA Network Open、Briefings in Bioinformatics、Biometrics、Biostatistics和Statistics in Medicine等医学和生物统计权威期刊上发表论文100余篇。主要学术兼职：中华预防医学会生物统计学分会第一届青年委员会主任委员，中华预防医学会生物统计分会和中国卫生信息与健康医疗大数据学会统计理论与方法委员会常务委员、中国现场统计研究会多元分析应用专业委员会与全国工业统计学教学研究会健康医疗大数据学会常务理事，以及《中国卫生统计》、《中国预防医学》、《Biostatistics & Epidemiology》等杂志编委。