Introduction to Causal Inference for Time-Varying Treatments Felix Elwert, University of Wisconsin – Madison
Saturday November 10th, 2007, 9:30 am – 5:30 pm Yale Department of Sociology This one-day workshop provides a broad conceptual introduction to counterfactual causal inference for time-varying treatments as developed by Jamie Robins and colleagues in biostatistics. The workshop starts with a refresher lecture on the basics of the counterfactual approach (potential outcomes, fundamental problem of causal inference, experimental analogy, matching). Next, it introduces Judea Pearl’s directed acyclic graphs (DAG) as a powerful and remarkably intuitive tool for deriving non-parametric identification results for a wide range of causal questions. Finally, the workshop uses DAGs to explain the particular challenge of causal inference for time-varying treatments, the inadequacy of standard regression models, and intuition for one solution (marginal structural models and inverse probability of treatment weighting). The emphasis throughout the course is on intuition and conceptual understanding rather than mathematical derivations and how-to mechanical recipes. Participants should have a thorough understanding of linear and logit regression and some familiarity with the counterfactual approach to causal inference. No calculus or matrix algebra required. This workshop is open to graduate students, post-docs, and faculty but the number of places is restricted. Lunch will be provided. Anyone wishing to participate should register with Chelsea Rhodes (chelsea.rhodes@yale.edu) no later than Friday, November 2nd, 2007. Felix Elwert is assistant professor of sociology at the University of Wisconsin – Madison, and recently graduated from Harvard University with degrees in sociology and statistics. He works on causal inference in sociology and demography and his research interests include marital dissolution and the social transmission of health through networks
|
