B1297
Title: Causal ball screening: Outcome model-free variable selection for causal inference
Authors: Dingke Tang - University of Toronto (Canada) [presenting]
Dehan Kong - University of Toronto (Canada)
Wenliang Pan - Sun Yat-sen University (China)
Linbo Wang - University of Toronto (Canada)
Abstract: Causal inference has been increasingly reliant on observational studies with rich covariate information. To build tractable causal procedures, including the doubly robust estimation, it is imperative to first extract important features from high dimensional data. Unlike the familiar task of variable selection for prediction modelling, our feature selection procedure aims to control for confounding while maintaining efficiency in the resulting causal effect estimate. Previous empirical and theoretical studies imply that one should aim to include all predictors of the outcome, rather than the treatment, in the propensity score model. We formalize this intuition through rigorous proofs and propose the causal ball screening for selecting these variables from modern ultra-high dimensional data sets. A distinctive feature of our proposal is that our procedure is more efficient than existing methods while our procedure keeps the desirable double robustness property. Our theoretical analyses show that the proposed procedure enjoys a number of favorable properties, including model selection consistency, normality and efficiency. Synthetic and real data analyses show that our proposal performs favorably with existing methods in a range of realistic settings.
