Causal Inference
Most scientific questions of interest in health sciences are causal in nature. For
example, does the intervention reduce risk of death in the target population? Randomized
controlled trial is the gold standard for inferring causal relationships. When randomization
is infeasible, researchers resort to identifying causal effect from observational
data such as sample surveys, censuses, and administrative records. Faculty in Biostatistics
are actively involved in the design, conduct, and analysis of both clinical trials
and observational studies, and have made fundamental contributions to the methodology
development for causal inference. Topics include experimental design, propensity score
methods, dynamic treatment regimes, missing data, unmeasured confounding, nonparametric
and semiparametric estimation, analysis of routinely collected administrative healthcare
data such as electronic health records.
Faculty: M. Banerjee, P. Han, R. Ladhania, J. Morrison, Xu Shi, J. Taylor, L. Wang, X. Zhou