Motivated by a real data example on renal graft failure, we propose a new semiparametric multivariate joint model that relates multiple longitudinal outcomes to a time-to-event. To allow for greater flexibility, key components of the model are modelled nonparametrically. In particular, for the subject-specific longitudinal evolutions we use a spline-based approach, the baseline risk function is assumed piecewise constant, and the distribution of the latent terms is modelled using a Dirichlet Process prior formulation. Additionally, we discuss the choice of a suitable parameterization, from a practitioner's point of view, to relate the longitudinal process to the survival outcome. Specifically, we present three main families of parameterizations, discuss their features, and present tools to choose between them.

, , , , ,,
Statistics in Medicine
Erasmus MC: University Medical Center Rotterdam

Rizopoulos, D., & Ghosh, P. (2011). A Bayesian semiparametric multivariate joint model for multiple longitudinal outcomes and a time-to-event. Statistics in Medicine, 30(12), 1366–1380. doi:10.1002/sim.4205