报告题目:When mediation analysis faces subgroup heterogeneity
报告人： 黄超 (佛罗里达州立大学)
报告时间：2023年6月29日16:40-17:40
报告地点：文波楼201教室
报告摘要:Mediation analysis is an important tool in the imaging genetics study for Alzheimer's disease (AD), where the goal is to identify the causal mechanism or pathway that links the genetic exposures and neurological outcomes, through some neuroimaging mediators. Although various mediation analysis approaches have been proposed to discover the underlying causal pathway in AD, there are several challenges, such as the subgroup heterogeneities in terms of (i) brain connectomes and (ii) causal mechanisms. To address these issues, we propose a novel mediation analysis tool that can simultaneously detect individual brain connectomes and subgroup causal pathways. Specifically, a two-layer structure mixture model, including a mixture of conditional Gaussian graphical models, is developed to establish the heterogeneous mediation pathways. A penalized EM algorithm is proposed to estimate both the average direct effect and indirect effect. Both simulation studies and a real example using the diffusion tensor imaging data from the ADNI study are conducted to assess the finite sample performance of our method.
报告人简介：黄超，现任佛罗里达州立大学统计系助理教授，于2008年，2014年在东南大学分别获得应用数学学士和博士学位，2019年在美国北卡罗来纳大学教堂山分校获得生物统计博士学位，研究方向包括生物统计，医学图像分析，函数数据分析，形状数据分析等。在国际高水平统计学期刊Journal of the American Statistical Association，Biometrika等上发表学术论文40多篇。
