Machine Learning Approaches for Propensity Score Estimation With Semi-Continuous Exposure

• In Event: AERA Roundtable Session Thursday 1:45 pm Four Seasons Ballroom 4
  In Roundtable Session: Machine Learning Techniques II: Advancing Coding/Essay Scoring and Estimation of Treatment Effects (Table 2)

Abstract

This study aims to evaluate the effectiveness of machine learning methods for propensity score estimation in semi-continuous exposure contexts. Utilizing a Monte Carlo simulation, we will compare the performance of various machine learning techniques, including Gradient Boosting Machines, Neural Networks, Random Forests, SuperLearner, and Support Vector Machines, against traditional parametric methods. The simulation will assess these methods under different conditions, such as varying levels of covariates, sample sizes, treatment effect sizes, dispersion parameters, and zero proportions. The goal is to identify the most effective approaches for achieving covariate balance and accurate treatment effect estimates in zero-inflated datasets. Our findings will provide valuable guidelines for researchers dealing with complex data structures across various domains.

Author

• Huibin Zhang, University of Tennessee