CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Regularized Cox frailty models for time varying covariates Authors:  Maike Hohberg - University of Goettingen (Germany) [presenting]
Andreas Groll - Technical University Dortmund (Germany)
Abstract: A method is proposed to regularize Cox frailty models that accommodates time-varying covariates and is based in the full likelihood. In previous simulation studies, it has been shown that using the conventional partial likelihood compared to the full likelihood yields a loss in precision especially in small or moderate samples. Given that in many medical applications for example, the sample size is often rather small, it seems surprising that none of the established \texttt{R} routines are based on the full likelihood considering that for small datasets using the full likelihood does not drastically increase computing time. We provide a function \texttt{coxlasso} that fits regularized Cox models accommodating varying coefficients effects and changes in the covariates. We show the function's superior performance compared to existing routines and assess situations in which using the full likelihood might be most effective.