A Systematic Review of Machine Learning for Propensity Score Estimation

• In Event: Educational Statisticians SIG Paper Session

Abstract

Machine learning has become a common approach for estimating propensity scores for quasi-experimental research using matching, weighting or stratification on the propensity score. However, there have not been comprehensive guidelines for using machine learning for propensity score estimation. This systematic review examined applications across different fields, such as statistics, medicine, education, and social sciences. The results show that the gradient boosting machine (GBM) is the most frequently used method, followed by neural network and random forest. The review showed a wide range of hyperparameter configurations and propensity score method implementations. These findings provide valuable insights for researchers seeking to employ machine learning in propensity score estimation and facilitate evidence-based decision-making in various domains.

Authors

• Walter L. Leite, University of Florida

• Huibin Zhang, University of Tennessee

• Zachary K. Collier, University of Connecticut

• Kamal Chawla, University of Maine

• Lingchen Kong, University of Florida

• Yongseok Lee, Purdue University

• Jia Quan, University of Kansas

• Olushola Soyoye, University of Delaware