Practical Causal Analysis and Effect Estimation with Observational Data

Full day short course (Monday, July 19; 9:00AM-5:00PM US Eastern)

Because of limited resources and practical factors in doing randomized experiments, psychological research often needs to infer causality from observational data. How can you determine whether an effect estimate has a valid causal interpretation? This course introduces commonly used methods for estimating dichotomous treatment effects from observational data and describes graphical models for evaluating the conditions under which the effect estimate has a valid causal interpretation. Specifically, this course describes estimation methods such as propensity score matching, inverse probability weighting, and doubly robust methods. It overviews the regression methods for causal mediation analysis. In addition, it explains the role of directed graphs as a tool to represent the data generating process, reason about sources of association and bias, and construct a valid estimation strategy. After an overview of the theory behind these methods and tools, simulated real-world examples use them to demonstrate good practices and effective strategies for dealing with practical challenges. This course emphasizes analytical techniques rather than the SAS software that is used in the examples; knowledge learned from this course can be used with other software. This is an introductory course―no prior experience with causal modeling is required.

About the Instructor

Yiu-Fai Yung

Yiu-Fai Yung is a senior manager in Advanced Analytics R&D at SAS Institute, where he develops statistical software for causal analysis and structural equation modeling. Most recently, he implemented the CAUSALMED procedure for causal mediation analysis. He obtained his PhD in psychology and MA degree in math from UCLA. He has led several structural equation modeling and causal analysis workshops at conferences such as SAS user group meetings and the Joint Statistical Meetings. Before joining SAS, he taught psychological statistics at the University of North Carolina at Chapel Hill.