|
Abstract:
|
It is well known that the choice of geographic unit can affect the interpretation of maps and the results of spatial analysis, a phenomenon known as the modifiable areal unit problem. A model-based hierarchical structure is developed to analyze data whose geographic units are misaligned in both space and time. We extend the multivariate Gaussian models to generalized linear mixed models that incorporate several forms of discrete data. The approach is illustrated with a data set relating illicit drug use, a range of sociodemographic variables, geography, and time in Tracy, CA. Here geography is indicated with ZIP codes which change over a period of eleven years. Computing is implemented via a carefully tailored Metropolis-Hastings algorithm, with map summaries created using a geographic information system.
|
