Informatik, TU Wien

Covariate Selection for Causal Inference with Observational Data

The unbiased estimation of causal treatment effects from observational data requires a statistical analysis that conditions on all confounding covariates.

Abstract

The unbiased estimation of causal treatment effects from observational data requires a statistical analysis that conditions on all confounding covariates. All the confounding bias can only be removed if the selection mechanism is ignorable, that is, if all confounders of treatment selection and the outcome are available and reliably measured. Ideally, covariates are selected according to a well-grounded substantive theory about the selection process and the outcome-generating model. However, with weak or no substantive theories, covariate selection strategies become more heuristic. In my talk I briefly introduce the “Rubin Causal Model”, discuss classes of bias-inducing and bias-amplifying covariates, and outline different strategies for selecting covariates in practice. Implications for cluster data (multilevel data) will also be addressed.

Biography

Peter M. Steiner is an assistant professor in Educational Psychology at the University of Wisconsin—Madison. His main research interest is in the methodology of causal inference, particularly quasi-experimental designs using propensity score methods, regression discontinuity designs, and interrupted times series designs. The methodology of factorial surveys (vignette experiments) is another field of research interests. His publications include contributions to the Journal of the American Statistical Association, Multivariate Behavioral Statistics, Psychological Methods, Journal of Educational and Behavioral Statistics, and Methodology: European Journal of Research Methods for the Behavioral and Social Sciences.

Note

This talk is organized by the Vienna PhD School of Informatics.