Resumen de Multivariate bayesian latent structure modeling of spatio-temporal health data
Andrew B. Lawson, Daniel R. Baer, Jane E. Joseph
Spatio-temporal health data is now routinely available. Often when time augments space, the focus is on modeling global spatio-temporal e ects.
However, temporal e ects are often localized spatially and so it could be important to disaggregate these e ects. This leads to spatial clustering of temporal e ects. Often this disaggregation is approached via latent mixture component models. Extending this approach to multiple disease incidence is the focus of this presentation. The speci c example that is explored, and motivates the detailed modeling, is incidence of mild cognitive impairment (MCI) and Alzheimers disease (AD). MCI is considered a pre-cursor of AD and so there is a temporal latent link between these outcomes. Our models address latent component mixtures for each disease but also coupled components shared between diseases. A case study in annual county level incidence in South Carolina is presented.
