Penalized Ridge Regression for Error‐Prone Additive Hazards Models

In this paper, we develop a penalized ridge regression approach for the additive hazards model in the presence of measurement error, collinearity, and uninformative covariates. These challenges frequently arise in survival data with multivariate covariates but are rarely addressed simultaneously. Specifically, we propose to minimize the error‐corrected deviance with constraints, which is used to control variable selection. For the numerical computation, we adopt the Newton‐barrier iteration to solve the constrained minimization problem. Theoretically, we establish some asymptotic properties, including variable selection consistency and asymptotic normality of the proposed estimator. Simulation studies under various magnitudes of measurement error, collinearity, and censoring rates demonstrate the validity of the proposed method and verify the stability and efficiency of the estimator. We further illustrate its practical utility through an application to the ACTG clinical trial data.