【专家简介】：宋学坤，密歇根大学安娜堡分校的生物统计学教授。他于1996年在加拿大不列颠哥伦比亚大学(温哥华)获得统计学博士学位。已发表超过230篇论文，培养了26名博士研究生，并指导了6名博士后研究员。他的研究兴趣包括数据整合、分布式推断、高维数据分析、纵向数据分析、中介效应分析以及时空建模。他是国际数理统计学会(IMS)、美国统计学会(ASA)Fellow以及国际统计学会(ISl) elected member。目前，他担任Journal of American Statistical Association、The Annals of Applied Statistics以及Journal of Multivariate Analysis的副主编。
【报告摘要】：Mediation pathway analysis becomes rather complex when areal data are spatially correlated. This is because interference from neighboring locations emerges to produce spillover effects on outcomes of a target location. To assess such spillover effects, we proposed an extension of exposure mapping via the Spatial Structural Equation Model (RES-SEM). This modeling approach enables us to rigorously estimate how, and to what extent, neighboring exposures and mediators may influence outcomes at a target location. We establish key identifiability conditions for causal mediation and interference estimands which are estimated by the maximum likelihood estimation (MLE)established using the no-i.i.d. Theoretical guarantees are asymptotic theory. Our methodology is applied to analyze a county-level infectious disease model, examining the causal effect of political party affiliation on COVlD-19 mortality, mediated by vaccination hesitancy and compliance.
【报告时间】：2024年12月16日 15:00-16:00
【报告地点】：崇真楼110
