Modeling Multiple Time-Varying Related Groups: A Dynamic Hierarchical Bayesian Approach With an Application to the Health and Retirement Study

As the population of older individuals continues to grow, it is important study relationship among variables measuring financial health and physical better understand demand for healthcare, insurance. We propose a semiparametric approach jointly model these variables. use data from Health Retirement Study which includes set correlated longitudinal health. In particular, we dynamic hierarchical matrix stick-breaking process prior some parameters account time dependent aspects our data. This introduces dependence across different groups varies over time. A Lasso type shrinkage specified covariates with time-invariant effects selecting significant on outcomes. Through joint modeling, are able conditional their health, vice-versa. Based analysis, find that insurance (medicare) provided by government (of United States) very effective, covers most medical expenditures. However, none insurances conveniently cover additional expenses due chronic diseases like cancer heart problem. Simulation studies performed assess operating characteristics proposed modeling approach. Supplementary materials this article, including standardized description available reproducing work, as an online supplement.