"Controversies in Modern Mediation Analysis"

Over the last decade there has been considerable development of new methods and wider substantive application of mediation analysis. Mediating variables are important for theoretical and applied research in many research areas because they are used to investigate how two variables are related. Examples are the process by which an intervention changes behavior and how a risk factor leads to disease. The purpose of this presentation is to give an overview of controversies in mediation analysis with attention to solutions for the limitations of these methods from recent work on causal inference, longitudinal models, and statistical testing.