Course format: Four hours of lecture/discussion per week
Credit: 4 units
Time: Mondays and Wednesdays 10am-12pm
Location: 2301 Tolman hall Room change
Instructors: Kathleen Mortimer and Romain Neugebauer
Email address: ph252d@lists.berkeley.edu

For directions to Tolman hall, please consult the campus map.

Course description: Marginal Structural Models (MSMs) are introduced as alternative statistical tools for causal inference in Epidemiology and compared to the conventional regression models for both point treatment and longitudinal data from observational studies. Estimation of causal effects defined by MSMs is related to estimation with missing data and developed with a focus on the Inverse Probability of Treatment Weighted (IPTW) estimator. Emphasis is placed on the assumptions on which causal inference with MSMs relies. Common concepts like confounding, selection bias, randomization, model specification, ... are revisited based on causal modeling with Directed Acyclic Graphs (DAGs) and the counterfactual framework. The real-life importance and applications of these abstract concepts are illustrated based on programming assignments where students simulate and analyze concrete data structures with the R software. Students are exposed to introductory applications of MSMs and IPTW estimation with programming assignments and a final class project. Students also review published methodological and subject-matter literature to evaluate statistical methods relative to MSMs. Performance evaluations are based on weekly assignments, class participation, a mid-term quiz and a final class project.

The course material (syllabus, slides and assignments) for this class is available with a password from the instructors.