S681: Statistical Methods For Causal Inference
Correlation is not causation. The quest for causation has formed a main stream in contemporary statistics. Based on the potential outcomes framework, this course presents the state-of-art of statistical methods for causal inference. The topics to be covered include inference in randomized experiments, matching on covariates, propensity score methods, directed acyclic graphs, instrumental variable methods, regression discontinuity designs, causal inference in panel data, causal mediation analysis, causal inference under interference, etc. A variety of examples are drawn from social and medical sciences to illustrate the methods.
Semester(s) Offered: Spring
Class time: Tuesday, Thursday : 11:15-12:30PM
Other Contact(s): Weihua An - email@example.com
Prerequisites: A basic understanding of logistic regression.