On January 5 from 4:30-5:30, Xu Qin will share her teaching demonstration titled “An Introduction to Causal Mediation Analysis”.
The lecture is designed to help students gain a conceptual understanding of how to define and identify causal effects in mediation studies. Many important research questions in education relate to how interventions work. Mediation analysis provides valuable information about the underlying mechanism through which the intervention generates a causal effect on the outcome. Traditionally in social sciences, mediation analysis has been formulated within the framework of linear structural equation models. However, this line of research has been challenged, due to the lack of a general definition of causal mediation effects, the void of a clarification of key identification assumptions, and the reliance on correct functional or distributional forms. In the past two decades, counterfactual thinking has given causal mediation analysis a sound theoretical basis. Under the counterfactual causal framework, I will demonstrate the definitions of “indirect effect” that channels the intervention impact through the mediator and “direct effect” that works through other unspecified mechanisms. The definitions involve potential outcomes and potential mediators associated with counterfactual treatment conditions. A major challenge in causal inference is that these quantities are unobserved. I will then clarify the identification assumptions, under which the unobservable quantities can be related to the observed data. This will allow us to provide the mediation effects with a causal interpretation and make valid inferences about the causal mediation effects as defined.