Primary Submission Category: Machine Learning and Causal Inference
Propensity Score Estimation for Generalized Treatment Spaces
Authors: Alessandro Leite, Felipe Lourenço Angelim Vieira, Alessandro Leite,
Presenting Author: *
Propensity scores comprise an effective strategy to control for biases when estimating causal effects based on observational data. Nevertheless, traditional methods often struggle with complex treatment structures beyond binary or discrete treatments. This work introduces a methodology for propensity score estimation that accommodates generalized treatment structures without imposing restrictive parametric assumptions. We follow a probabilistic classification model to estimate the stabilized weights and the propensity scores via density ratio estimation, making our method adaptable to various treatment forms. Experimental results show notable performance improvements in accuracy and stability across different synthetic and semi-synthetic datasets, even when the number of covariates increases. Such results suggest that our approach is flexible enough for challenging real-world applications when estimating causal effects such as healthcare, policy analysis, and personalized marketing.