|
Abstract:
|
Correlated longitudinal data collected from studies that focus on the well-being of patients and their family members are usually complex. The key features include multiple outcomes of interest collected longitudinally from multiple members within families and the interrelationships of these outcomes over time. An approach that can properly analyze such complex interconnected data is of interest in practice. I develop a Bayesian semi-parametric approach that combines longitudinal bivariate regressions with actor-partner interdependence models through shared random effects. The proposed model uses a non-parametric prior, i.e., Dirichlet process prior, for the shared random effects to relax the normality assumption. This approach allows researchers to assess the trajectories of multiple outcomes for patients and their family members jointly in a single hierarchical model, to account for clustering within individuals and within families, and to investigate the associations of these outcomes between the correlated populations over time. A pilot study will be used to demonstrate the proposed approach.
|
