In Fall 1999 , we began an introduction to the study of causal inference based on counterfactuals (Click here for a description of this topic and abstracts and slides from past talks). The notion of making causal statements on the basis of observed data has become an area of great current interest, as this idea is to many statisticians inconsistent with the historically held perspective that only associational inferences may be drawn from observational data (including randomized experiments that do not turn out as planned). Interest in making causal inference reaches beyond just biomedical and epidemiological applications; questions of casuality are the focus in research in applications such as in economics, social sciences, education, political science, philosophy, and artifical intelligence. Numerous sessions on this topic were held at the ENAR meetings in March 2000 and at the Joint Statistical Meetings in August 2000, and a short course was presented at ENAR as well. A conference on the topic bringing together researchers from numerous disciplines as well as statisticians will be held in August 2001 immediately prior to the Joint Statistical Meetings.
Seminars and copies of slides, Spring 2000:
Marginal structural models and inverse-probability-of-treatment weighted estimators (2/11/00)
The use of Robins' marginal structural models for causal inference (2/24/00)
( Slides in PostScript )
Estimating the causal survival effect between two treatments in the presence of confounding and censoring (3/9/00)
Causal inference and the "null paradox" for time-varying treatments (3/30/00)
( Slides in PostScript )
Structural nested models for causal inference (5/4/00)
Return to Biostatistics Working Group