Statistics Seminar: Beyond the Average Treatment Effect: Causal Mediation Methods for Understanding Intervention Mechanisms
Presenter: Hanna Kim, Assistant Professor, Psychology Department, UCSC
Description: Understanding how an intervention works is a central question in behavioral and social research, following the demonstration of its overall effect. Traditional mediation analysis techniques often assume a homogeneous mechanism of effects, overlooking both validity concerns and subgroup variation in causal pathways. In this talk, I present a series of developments in causal mediation methods aimed at identifying and estimating natural direct and indirect effects, addressing challenges such as unobserved confounding and heterogeneity across subpopulations. I illustrate these approaches using data from early childhood education programs and online course participation, showing how causal mediation analysis can reveal diverse mechanisms of change. The talk concludes with current directions for integrating mixture modeling and clustered data analysis with causal inference to enhance both robustness and interpretability at the interface of statistics and applied research.
Bio: Hanna Kim is an Assistant Professor of Quantitative Psychology at the University of California, Santa Cruz. Her research centers on advancing causal inference and mediation methods to investigate how educational programs influence child development and how effects differ across subpopulations. She also integrates latent variable modeling with dyadic data analysis to study interpersonal dynamics. Her work bridges psychology and statistics to address methodological challenges in applied research, with publications in Psychological Methods, Cancer Epidemiology, Biomarkers & Prevention, Journal of Substance Use and Addiction Treatment, and the Asian Journal of Education.
