Abstract: In this article, we carefully examine two important implementation issues when estimating propensity scores using generalized boosted models (GBM), a promising machine learning technique. First, we examine which of the following methods for tuning GBM lead to better covariate balance...