Propensity Score Method with Joint Modeling of Mixed-Type Treatment Variables
Author
Grace Chiu
Co-author
-
F. Swen Kuh
-
Anton H. Westveld
-
-
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: bayesian modeling, causal treatment effect, copulas, observational studies
Session: IPS 894 - Advancements in Statistical Methodologies for Environmental and Health Data Analysis
Thursday 9 October 10:50 a.m. - 12:30 p.m. (Europe/Amsterdam)
Abstract
Current propensity score (PS) modeling frameworks for causal inference in observational studies do not readily allow for simultaneous handling of mixed-type treatment variables, such as continuous, discrete, and ordinal variables. We propose an extended rank likelihood method that utilizes copulas to model latent propensity scores for mixed-type treatment variables. We illustrate two versions of our method on real-world datasets and compare their performance against the traditional ("non-latent") PS method.