On the Uniform Frailty Model with Penalized Likelihood and Clustered Data

In the field of survival analysis, when heterogeneity is suspected across study subjects, a model that can account for that variability is preferred. Moreover, an important and challenging task in that field is to efficiently select a subset of significant variables upon which the hazard function depends. To this end, frailty models along with the penalized likelihood methodology can be applied. In this paper, we extend the Gamma frailty model methodology of Fan and Li (2002) to the Uniform frailty model, based on the results of our previous work (Androulakis et al., 2012). Theoretical findings are illustrated via a thorough simulation study.