Bridging Theory, Data, and Method: A Framework for Modeling Heterogeneous Treatment Effects in Educational Research

• In Event: AERA Poster Session Sunday 1:45 pm
  In Poster Session: Methodological Techniques for Challenging Data and Modeling Situations

Abstract

Valid educational evaluation requires asking not only whether a program works, but for whom, under what conditions, and how effectively. While average treatment effects offer useful summaries, they often obscure critical variation. This paper presents a structured approach for modeling heterogeneous treatment effects, integrating a taxonomy of heterogeneity types with analytic guidance. Drawing on causal inference theory and statistical modeling principles, we link theoretical questions with appropriate estimands and estimation strategies. We illustrate the approach using empirical applications and highlight how modeling decisions can enhance the rigor, interpretability, and policy relevance of educational research. This structure helps researchers design and interpret studies that capture variation across learners, contexts, and outcomes, supporting both theoretical refinement and practical decision making.

Authors

• Jee-Seon Kim, University of Wisconsin - Madison

• Graham Ward Buhrman, University of Wisconsin - Madison

• Xiangyi Liao, University of British Columbia