Title: Outcome Based Matching

Abstract: We propose a method to reduce variance in treatment effect estimates in the setting of high-dimensional data. In particular, we introduce an approach for learning a metric to be used in matching treatment and control groups. The metric reduces variance in treatment effect estimates by weighting covariates related to the outcome and filtering out unrelated covariates.