Causal Inference: Theory and Practice
BIOSTAT 258
Subject & Catalog Number
Course Information
Description
Randomized experimentation is the standard for quantifying the causal effect of an intervention in the public health and biomedical sciences, yet randomization may be impossible, impractical, or unethical, leading to real-world scenarios in which causal inferences are based upon observational comparisons. This course reviews the foundations of causal inference in (bio)statistics, outlining causal-analytic methods that help to extract as much evidence as imperfect observational studies carry about causal effects commonly of interest in applied health science settings. This survey of statistical causal inference begins with foundational concepts: the potential outcomes framework and counterfactual random variables, graphical modeling frameworks, and common assumptions and strategies for the identification of the causal effects of static and dynamic interventions. Methodological extensions for studying the causal effects of time-varying, longitudinal interventions will be touched upon as well. Further topics to be addressed may include instrumental variables approaches, principal stratification, causal mediation analysis, treatment effect heterogeneity, the causal dose-response curve, and causal survival analysis. This course may also introduce elements of semi-parametric theory necessary for the development of asymptotically efficient estimators of causal effect estimands. Where possible, the role of modern regression (i.e., machine learning) techniques and tools for the practical estimation of causal effect estimands will be emphasized.
Course Notes
Pre-requisites: BIOSTAT 231 and BIOSTAT 232.
Available for Harvard Cross Registration
