Treatment Effect Performance of the X-Learner in the Presence of Confounding and Non-Linearity

摘要

This study critically evaluates a recent machine learning method called the X-Learner,that aims to estimate treatment effects by predicting counterfactual quantities. It uses informationfrom the treated group to predict counterfactuals for the control group and vice versa. Theproblem is that studies have either only been applied to real world data without knowing theground truth treatment effects, or have not been compared with the traditional regression methodsfor estimating treatment effects. This study therefore critically evaluates this method by simulatingvarious scenarios that include observed confounding and non-linearity in the data. Althoughthe regression X-Learner performs just as well as the traditional regression model, the other baselearners performed worse. Additionally, when non-linearity was introduced into the data, theresults of the X-Learner became inaccurate.