Week 6

A. Solomon Kurz - Professor, The Chicago School of Professional Psychology

Causal Inference with Real-World Data from Randomized Experiments

Randomized experiments make it relatively easy for us to make valid causal inferences. When you have access to them, baseline covariates can help improve the quality of our causal inferences. Though baseline covariates are relatively easy to use with OLS-style regression, they can be surprisingly difficult to use in many other kinds of regression (e.g., logistic regression). We can overcome this difficulty with what is sometimes called the "standardization" or "g-computation" approach. This talk will provide an overview of how this all works.