A Nonseparable Multivariate Space-Time Model for Analyzing County-Level Heart Disease Death Rates

• Activity Number: 644
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Harrison Quick* and Lance Waller
• Companies: CDC and Emory University
• Keywords: Bayesian methods ; Gender disparities in health ; Racial disparities in health ; Spatio-temporal data analysis

Abstract

While death rates due to heart disease have experienced a sharp decline over the past 50 years, these diseases continue to be the leading cause of death in the United States. Here, we look to harness the power of hierarchical Bayesian methods to analyze a dataset comprised of county-level, temporally varying heart disease death rates for men and women of different races from the US. Specifically, we propose a nonseparable multivariate spatio-temporal Bayesian model which not only allows for group-specific temporal correlations, but also allows for temporally-evolving covariance structures in the multivariate spatio-temporal component of the model. Furthermore, the model is capable of seamlessly handling counties with missing data due to the lack of observations from a particular subpopulation. After verifying the effectiveness of our model via simulation, we apply our model to our dataset of over 200,000 county-level heart disease death rates. In addition to yielding a superior fit than other common approaches for handling such data, the richness of our model provides deeper insight into racial, gender, and geographic disparities underlying heart disease death rates in the US.