A0773
Title: High dimensional recurrent event analysis with error-contaminated covariates
Authors: Kaida Cai - Southeast University (China) [presenting]
Abstract: In the analysis of recurrent event data with high-dimensional covariates, measurement error in covariates poses significant challenges for reliable variable selection and risk estimation. Naively ignoring measurement errors often leads to biased and misleading results. To address this issue, variable selection is considered for recurrent event data with high-dimensional covariates contaminated by measurement error. A penalized corrected likelihood approach is proposed to simultaneously adjust for measurement error and perform variable selection. The method is based on a corrected score function that accounts for the measurement error through an additive error model and a piecewise constant baseline hazard function. The theoretical properties of the proposed estimator are established, including consistency and the oracle property. Simulation studies demonstrate that the method yields improved selection accuracy and reduced false discovery rate compared to naive approaches. The utility of the method is further illustrated using a real dataset on recurrent events.
