A1390
Title: Quantile recurrence time model with time-dependent covariates
Authors: Peiming Lai - The Chinese University of Hong Kong (Hong Kong) [presenting]
Tony Sit - The Chinese University of Hong Kong (Hong Kong)
Abstract: Quantile regression offers a flexible framework for analyzing recurrent event data by modeling the time to a given expected frequency. Existing methods are, however, generally restricted to time-independent covariates, limiting their use in applications where dynamic predictors are of primary interest. A generalized quantile recurrence time model that accommodates time-dependent covariates under right-censoring is proposed, which subsumes conventional accelerated recurrence time models and establishes connections to censored quantile regression methods for survival analysis. A sequential estimation procedure based on a martingale-like estimating equation is developed, and the uniform consistency and weak convergence of the proposed estimator are established. A computationally efficient sample-based covariance estimator is provided for inference. Simulation studies demonstrate strong finite-sample performance, and an analysis of recurrent pharyngitis data illustrates the practical utility of the model in revealing dynamic covariate effects.
