On Model Selection and Model Misspecification in Causal Inference

46 Pages Posted: 23 Nov 2010

See all articles by Stijn Vansteelandt

Stijn Vansteelandt

Ghent University.Faculty of Scoiences

Maarten Bekaert

Ghent University.Faculty of Sciences. Department of Applied Mathematics and Computer Sciences

Gerda Claeskens

KU Leuven - Department of Economics

Date Written: 2010

Abstract

Standard variable-selection procedures, primarily developed for the construction of outcome prediction models, are routinely applied when assessing exposure e®ects in observational studies. We argue that this tradition is sub-optimal and prone to yield bias in exposure effect estimates as well as their corresponding uncertainty estimates. We weigh the pros and cons of confounder-selection procedures and propose a procedure directly targeting the quality of the exposure effect estimator. We further demonstrate that certain strategies for inferring causal effects have the desirable features (a) of producing (approximately) valid confidence intervals, even when the confounder-selection process is ignored, and (b) of being robust against certain forms of misspecification of the association of confounders with both exposure and outcome.

Keywords: Causal inference, Confounder selection, Double robustness, Influential weights, Model selection, Model uncertainty, Propensity score

Suggested Citation

Vansteelandt, Stijn and Bekaert, Maarten and Claeskens, Gerda, On Model Selection and Model Misspecification in Causal Inference (2010). Available at SSRN: https://ssrn.com/abstract=1713126 or http://dx.doi.org/10.2139/ssrn.1713126

Stijn Vansteelandt (Contact Author)

Ghent University.Faculty of Scoiences ( email )

Gent, 9000
Belgium

Maarten Bekaert

Ghent University.Faculty of Sciences. Department of Applied Mathematics and Computer Sciences ( email )

Gent, 9000
Belgium

Gerda Claeskens

KU Leuven - Department of Economics ( email )

Leuven, B-3000
Belgium

Register to save articles to
your library

Register

Paper statistics

Downloads
59
Abstract Views
504
rank
366,071
PlumX Metrics