- Bayesian Anal.
- Advance publication (2018), 24 pages.
High-Dimensional Confounding Adjustment Using Continuous Spike and Slab Priors
In observational studies, estimation of a causal effect of a treatment on an outcome relies on proper adjustment for confounding. If the number of the potential confounders () is larger than the number of observations (), then direct control for all potential confounders is infeasible. Existing approaches for dimension reduction and penalization are generally aimed at predicting the outcome, and are less suited for estimation of causal effects. Under standard penalization approaches (e.g. Lasso), if a variable is strongly associated with the treatment but weakly with the outcome , the coefficient will be shrunk towards zero thus leading to confounding bias. Under the assumption of a linear model for the outcome and sparsity, we propose continuous spike and slab priors on the regression coefficients corresponding to the potential confounders . Specifically, we introduce a prior distribution that does not heavily shrink to zero the coefficients (s) of the s that are strongly associated with but weakly associated with . We compare our proposed approach to several state of the art methods proposed in the literature. Our proposed approach has the following features: 1) it reduces confounding bias in high dimensional settings; 2) it shrinks towards zero coefficients of instrumental variables; and 3) it achieves good coverages even in small sample sizes. We apply our approach to the National Health and Nutrition Examination Survey (NHANES) data to estimate the causal effects of persistent pesticide exposure on triglyceride levels.
Bayesian Anal., Advance publication (2018), 24 pages.
First available in Project Euclid: 10 November 2018
Permanent link to this document
Digital Object Identifier
Antonelli, Joseph; Parmigiani, Giovanni; Dominici, Francesca. High-Dimensional Confounding Adjustment Using Continuous Spike and Slab Priors. Bayesian Anal., advance publication, 10 November 2018. doi:10.1214/18-BA1131. https://projecteuclid.org/euclid.ba/1541818861
- Supplementary materials for “High-dimensional confounding adjustment using continuous spike and slab priors”. Here we give additional details and derivations for estimation of the empirical Bayes variance and posterior calculation. We further illustrate the estimation of the posterior mode of our model and give additional simulation results. An R package implementing the approach for both binary and continuous outcomes is available at github.com/jantonelli111/HDconfounding.