Overview
Paul R. Rosenbaum is Professor Emeritus of Statistics at the Wharton School, University of Pennsylvania. He is one of the leading methodologists in the design and analysis of observational studies, with foundational contributions to propensity score methods (with Rubin), sensitivity analysis for hidden bias, and nonparametric inference for causal effects. His monograph Observational Studies (Springer, 1995; 2nd ed. 2002) is the standard reference for the statistical analysis of non-randomized studies.
Key Contributions / Features
- Propensity score — with Rubin (Biometrika 1983), showed that conditioning on the propensity score P(T=1|X) removes observed confounding; the foundational paper for matching and weighting estimators in observational studies
- Sensitivity analysis for hidden bias — developed a formal framework for assessing how strong an unmeasured confounder would need to be to explain away an estimated treatment effect; uses odds ratio Γ as the sensitivity parameter; extended by Rosenbaum (1993) to Hodges-Lehmann point estimates
- Sensitivity to nonignorable instrument assignment — extended AIR's sensitivity analysis to cover violations of instrument ignorability, computing how far the instrument assignment could deviate from random before estimated effects become unreliable; discussed in the JASA 1996 exchange (Rejoinder pp. 471–472)
- Hodges-Lehmann estimator for IV — proposed using rank-based robust estimators (Hodges-Lehmann) as alternatives to mean-based IV estimators; addresses the lack of robustness to outliers in standard IV moment estimators
- Weak exclusion restriction — pointed out that the exclusion restriction for non-compliers need not require defining counterfactual potential outcomes Y(1) for never-takers or Y(0) for always-takers; suffices to assume Y(0,D(0)) = Y(1,D(1)) for units with D(0)=D(1)
- Observational Studies (1995, Springer-Verlag) — standard reference for the statistical analysis of non-randomized data; covers matching, sensitivity analysis, and inference for causal effects
Related
Sources