Recently (February 11, to be exact), Jamie Robins presented the Department of Statistics seminar on his "Marginal structural models and inverse-probability-of-treatment-weighted estimators" (which he also presented the day before in the Department of Epidemiology at UNC-Chapel Hill). In this talk, I will review these models, the assumptions behind them, and inverse-probability-of-treatment-weighted (IPTW) estimators both in the point exposure (non-longitudinal) setting and the time-dependent-treatment setting and provide a heuristic exhibition of why the IPTW estimator yields consistent estimators of causal effects. I will also discuss Robins' "G-computation algorithm formula," which both provides an alternative way to estimate causal effects and gives justification of the fact that, under appropriate assumptions, causal effects are indeed identifiable from observed data on time-dependent-treatments. If you missed Jamie's talk, or were intrigued and would like another chance to see this material, this is the talk for you!
Return to Biostatistics Working Group