A0415
Title: Multivariate binary extension for W\&A-learner
Authors: Shintaro Yuki - Doshisha University (Japan) [presenting]
Kensuke Tanioka - Graduate School of Doshisha University (Japan)
Hiroshi Yadohisa - Doshisha University (Japan)
Abstract: Randomized controlled trials and observational studies are conducted to test the efficacy of a treatment, and the focus is on two-arm comparisons. The results of a treatment may not demonstrate efficacy in a population that meets the eligibility criteria. In such cases, it is desirable to efficiently identify populations with characteristics that make the treatment effective so-called subgroups. They can be identified by estimating the treatment effect. A prior study introduced the W-learner and A-learner as approaches for modeling the interaction between treatment modalities and covariates using propensity score weighting aimed at estimating the effect of treatments on a single outcome. Subsequently, a recent study extended the W-learner to handle multivariate outcomes. However, while this method accounts for the correlation structure among multiple continuous outcomes, it fails to address the correlation structure among multiple binary outcomes, and no multivariate extension has been applied to the A-learner. A novel approach that enables both the W-learner and A-learner to account for the correlation structure among multiple binary outcomes is proposed.
