报告题目：Two-fold Varying-coefficient Mediation Models and Their Applications

报 告 人：王文武教授（曲阜师范大学）

报告时间：6月24日16：00--17：00

报告地点：数学研究中心528

报告摘要：

Mediation analysis provides a technical approach to reveal the complex relationships between the variables, in such a way to distinguish more targeted interventions for different individuals and different scenario conditions. In real applications, the strength of causal effect between the variables is often influenced by many other variables. Yet in the literature, there are few studies on multiple mediation analysis to explore in this direction. Inspired by this, we study the two-fold varying-coefficient mediation model by proposing two influence mechanisms involving two mediators. Specifically, the first mechanism considers a parallel two-fold varying-coefficient mediation model where the two mediators operate in parallel, and the second one considers a chain two-fold varying-coefficient mediation model where the two mediators function sequentially. Further by regarding the coefficient functions as smooth functions, we derive the spline estimators for both the direct and indirect effects, as well as establish the asymptotic normality of all or part of the estimators. Simulation results show that our new models and methods perform very well, and they are also able to capture the dynamic changes of the coefficient functions in the two real data examples.

报告人简介：

王文武，曲阜师范大学统计与数据科学学院教授、博导；现任校统计学研究所副所长，院研究生工作负责人；香港大学博士后研究员，香港浸会大学访问学者。研究兴趣包含：非参数统计、稳健统计、机器学习与人工智能、因果效应评估等。主持国家自然科学基金面上项目1项、山东省自然科学基金面上项目1项。在Journal of Machine Learning Research、CSDA、Journal of Hazardous Materials、Knowledge-Based Systems、Psychometrika、Statistics in Medicine、Structural Equation Modeling、TEST、《统计研究》等学术期刊发表论文20余篇。