A0734
Title: A joint model with finite-mixture structure for longitudinal and survival data
Authors: Xuerui Yang - University of Manchester (United Kingdom) [presenting]
Jianxin Pan - The University of Manchester (United Kingdom)
Christiana Charalambous - University of Manchester (United Kingdom)
Abstract: In longitudinal studies, heterogeneous time-to-event data is commonly collected together with longitudinal data. A finite-mixture Cox PH model is used to quantify the probabilities of clusters in survival data, and joint mean-covariance models are used to characterize within-subject patterns in the longitudinal outcomes. These two sub-models are linked via a shared parameter, but a third sub-model is also introduced to show how the covariance structures are connected. An MCMC algorithm for parameter estimation is proposed, and the models' performance is demonstrated via simulation studies. An application to AIDS data illustrates how the joint models may be used to capture the heterogeneity and longitudinal patterns.
