ST 599 - W 25: Missing data and causal inference
Syllabus
Textbook:
Required: Statistical Analysis with Missing Data, 3rd edition Little and Rubin
Required: Causal Inference for Statistics, Social, and Biomedical Sciences, Imbens and Rubin
Optional: Bayesian Inference for Partially Identified Models, Gustafson
Course objectives
Upon completion of this course, students should be able to critically evaluate how published literature handles (or does not handle) missing data, and analyze datasets that have missing values by designing models that account for missingness. Students should also be able to read published literature using randomized study designs, and assess whether researchers’ causal conclusions are reasonable.
Course learning outcomes
- Differentiate between missing-completely-at-random, missing-at-random (MAR), and missing-not-at-random (MNAR) processes via assumptions about the joint distribution of missingness indicators, outcomes, and covariates.
- Evaluate whether estimands of interest are identifiable for a given data generating process.
- Derive the identification region and limiting posterior density for partially-identified models.
- Derive a principal causal effect using the Neyman-Rubin causal model.
- Construct and fit maximum likelihood (in R)/Bayesian models (in Stan) for MAR, MNAR, and causal models.