Using SAS software for Multilevel Models in Survival Analysis

Abstract

Multilevel models have been used to analyze nested data to take into account the dependence structure that may be present. This is accomplished by including random coefficients in the model. In this work, we consider the multilevel framework adapted to common situations in Survival Analysis. Two approaches are discussed, one based on a parametric model (Weibull), and other on the semi parametric Cox proportional hazards model. Assuming that the random coefficients are normally distributed, we discuss how to use GLIMMIX macro to fit both models. We also present a macro to optimize the procedure and show how to use it in a real data example, based on an experiment with rats.