Evaluating large-scale propensity score performance through real-world and synthetic data experiments

L1-regularization incorporates all covariates simultaneously into the propensity score model and offers propensity score performance superior to the hdPS marginal screen.