Special Seminar
Edward H. Kennedy, PhD Candidate
Research Assistant, Department of Biostatistics, University of Pennsylvania
Robust causal inference with continuous treatments
ALL ARE WELCOME
Abstract:
Continuous treatments (e.g., doses) arise often in practice, but available causal effect estimators are limited: they either require parametric models for the effect curve, or else do not allow for doubly robust covariate adjustment. This is an example of a more general problem: if we want to use standard semiparametric approaches then we must live with low-dimensional estimands, while if we want to pursue more complex estimands then typically we must rely on plug-in estimation (which is sensitive to the curse of dimensionality). To solve this problem in the continuous treatment effect setting, we develop a novel doubly robust kernel smoothing approach that requires only mild smoothness assumptions on the effect curve, and still allows for misspecification of either the treatment density or outcome regression (i.e., gives small second-order estimation bias). We derive asymptotic properties and propose a procedure for data-driven bandwidth selection. The methods are illustrated in a study of the effect of nurse staffing on hospital readmissions penalties.
Bio:
Edward Kennedy is a PhD student in biostatistics at the University of Pennsylvania.聽 His research interests include causal inference, high-dimensional data, and non- and semi-parametric methods, among other topics.聽 He is also particularly interested in applications in health services research, medicine and public policy.
Please visit: