Bayesian bivariate bent-cable model for longitudinal data

Abstract

Growth curve models are often used to describe a developmental course of a longitudinal response. This paper extends such models to assess multiphasic patterns of developmental trajectories for multivariate response variables. The multiphasic patterns are identified using a bivariate bent-cable model in the context of multivariate growth models. The approach allows for the simultaneous estimation of parameters of multiphasic changes in each response, and also takes into account the correlations among outcomes and random effects for repeated observations over time. The proposed methods are demonstrated using real data from an AIDS clinical study.