Fig. 2. Propensity score distributions for treatments with logistic, Tree, GBM, and SL algorithms. The Wald 95% CI is based on the robust SE. Each solid line and dotted line mean a group that has been treated and a group that has not been treated. Except for GLM model, machine learning models (TREE, GBM, and SL) show that the two graphs overlap quite well.
© 2022 Communications for Statistical Applications and Methods
