Academic Report of School of Mathematical Sciences [2026] No. 068

(Series Report for High-Level University Construction No. 1327)

Title：Semiparametric Efficient Mediation Analysis

Speaker：Wenwu Wang,  Professor (Qufu Normal  University)

Time：14:30-15:30, July 5, 2026

Location：Room 514, Huixing Building, Yuehai Campus, Shenzhen University

Abstract: Mediation analysis is a pivotal tool for inferring causal mechanisms in social science and related areas. Traditional ordinary least squares inference is highly sensitive to non-normal errors. Robust alternatives, like least absolute deviation and Huber methods, assume symmetric error distributions and often suffer significant efficiency loss. In this paper, we propose a semiparametric efficient indirect mediation effect estimator (SEIMEE) within the framework of maximum profile likelihood. Without requiring prior error assumptions, SEIMEE automatically captures the underlying data characteristics and asymptotically achieves oracle efficiency. Based on our new estimation framework, we propose the SEIMEE difference and product estimators for the indirect effect and establish the asymptotic distribution for the product estimator. A nonparametric percentile (PRCT) and a bias-corrected (BC) bootstrap procedure are developed to construct robust confidence intervals. To implement our kernel-based methodology, we propose a data-driven bandwidth selection strategy and a stable iterative alternating algorithm. Finite-sample simulation studies demonstrate that the proposed method exhibits superior performance in terms of mean squared error, type I error rates, and statistical power compared to existing methods in most cases. Finally, the effectiveness and practical utility of the proposed method are further validated through real data analysis.

Speaker Profile：Wenwu Wang is a professor at Qufu Normal University, a doctoral advisor, deputy director of the university’s Institute of Statistics, and head of graduate studies for his college. He has served as a postdoctoral researcher at the University of Hong Kong and a visiting scholar at Hong Kong Baptist University. His research interests include nonparametric statistics, robust statistics, machine learning and artificial intelligence, causal inference, and policy evaluation. Currently, he is leading one General Project from the Shandong Provincial Natural Science Foundation and one General Project from the Ministry of Education’s Humanities and Social Sciences Program; he has previously led and completed one General Project from the National Natural Science Foundation of China and one Key Project from the National Bureau of Statistics. He has published more than 20 SCI/SSCI/CSSCI papers, including in leading statistics journals such as the Journal of the American Statistical Association, Statistics in Medicine, TEST, and Computational Statistics & Data Analysis; leading machine learning and artificial intelligence journals such as Journal of Machine Learning Research and Knowledge-Based Systems; leading social science journals such as Structural Equation Modeling and Psychometrika; the top environmental statistics journal Journal of Hazardous Materials; and Chinese journals such as Statistical Research and Mathematical Statistics and Management.

Faculty and students are welcome to attend!

Invited by: Zongliang Hu

School of Mathematical Sciences

June 26, 2026