Models for Causal Inference – EPI289 (Intermediate), since 2004
Instructor: Miguel Hernán
EPI289 describes models for causal inference, their assumptions, and their practical application to epidemiologic data. The course introduces outcome regression, propensity score methods, the parametric g-formula, inverse probability weighting of marginal structural models, g-estimation of nested structural models, and instrumental variable methods. The course also introduces models for causal inference in the presence of time-varying exposures, which will be extensively studied in EPI207. EPI289 is designed to be taken after EPI201/EPI202. The epidemiologic concepts and methods studied in EPI201/202 will be reformulated within a modeling framework in EPI289. Familiarity with the SAS language is strongly recommended.
Spring Semester, Harvard T.H. Chan School of Public Health
Advanced Epidemiologic Methods – EPI207 (Advanced), since 2001
Instructors: James Robins and Miguel Hernán
This course provides an in depth investigation of statistical methods for drawing causal inferences from observational studies with time-varying treatments. Epidemiologic concepts such as time-varying confounding and selection bias, intermediate variables, overall effects and direct effects are formally defined within the context of a counterfactual causal model. Methods for the analysis of the causal effects of time-varying exposures in the presence of time-varying covariates that are simultaneously confounders and intermediate variables are emphasized. These methods include g-estimation of structural nested models, inverse probability weighting of marginal structural models, and the g-formula. As a practicum, students reanalyze data sets using the above methods.
Fall semester, Harvard T.H. Chan School of Public Health
Confounding control: a component of causal inference – EPI524 (Intermediate), since 2016
Instructors: Miguel Hernán and Sonja Swanson
Controlling for confounding is a fundamental component of epidemiologic research. This course describes models for confounding control (or adjustment), their application to epidemiologic data, and the assumptions required to endow the parameter estimates with a causal interpretation. The course introduces students to two broad sets of methods for confounding control: methods that require measuring and appropriately adjusting for confounders, and methods that do not require measuring the confounders. Specifically, the course introduces outcome regression, propensity score methods, the parametric g-formula, inverse probability weighting of marginal structural models, and instrumental variable methods as means for confounding control. EPI524 is designed to be taken after EPI522. The models described in EPI524 are for time-fixed dichotomous exposures and dichotomous, continuous, and failure time (e.g., survival) outcomes.
Spring semester, Master of Public Health in Epidemiology, Harvard T.H. Chan School of Public Health
Epidemiologic Methods I – EPI201 (Introductory), 2008-2012
Instructor: Miguel Hernán
EPI201 introduces epidemiologic methods for description and causal inference. The course discusses conceptual and practical issues encountered in the design, conduct, and analysis of epidemiologic studies. These fundamental concepts are studied in the simplified context of closed populations with time-fixed exposures, nonparametric methods for data analysis, and absence of sampling variability. The final exam requires the application of the learned skills to a real problem in epidemiology. EPI201 is the first course in the series of methods courses designed for students majoring in Epidemiology, and for those interested in a detailed introduction to epidemiologic methods. Students who take EPI201 are expected to take EPI202 (Methods II).
Fall Semester, Harvard T.H. Chan School of Public Health