A0448
Title: Rank-based inference for individualized treatment rules in single-index varying coefficient model
Authors: Yishan Cui - Indiana University Indianapolis (United States)
Honglang Wang - Indiana University Indianapolis (United States) [presenting]
Abstract: Individualized treatment rules (ITRs) provide critical guidance for patients by tailoring treatment decisions based on their specific covariates. However, deriving inferences for ITRs can be challenging, particularly when interactions between treatment and covariates are modeled non-parametrically, as this can introduce significant bias in estimating the ITRs. A unified rank-based inference procedure is proposed for ITRs under a semi-parametric single-index varying coefficient model, where the non-parametric coefficient function is assumed to be monotone increasing. To avoid direct estimation of the non-parametric function, the approach leverages maximum rank correlation. For hypothesis testing, the asymptotic distribution of the proposed estimator is not only derived using de-biasing techniques, but also the jackknife empirical likelihood is leveraged to test the significance of the treatment rule. The finite-sample performance of the proposed method is assessed through Monte Carlo simulations. The proposed method is further exemplified by its application to the ACTG175 data.
