Title: A joint mixed membership model for multivariate longitudinal and survival data
Authors: Yuyang He - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: In longitudinal studies, a conventional approach to capture the individual heterogeneity of the population is to adopt a finite mixture model by assigning each subject to a single cluster. However, some subjects may not exactly correspond to one typical cluster, but instead, behave somewhere in between two (or more) clusters. We propose a new joint mixed membership model to address such heterogeneity and investigate the relationship between multivariate longitudinal and survival data. A vector of probability weights for characterizing partial membership is introduced both on (i) a mixed-effects model for describing the trajectories of longitudinal observations and (ii) a PH model for examining the effects of time-dependent risk factors on the hazard of interest. We develop a Bayesian joint estimation method coupled with efficient Markov chain Monte Carlo sampling schemes to perform statistical inference for the new joint model. The proposed approach is assessed through extensive simulation studies and an application to Alzheimer's Disease Neuroimaging Initiative study.