Bayesian multivariate spatial models for roadway traffic crash mapping (Q2581521)

An extension of univariate spatial models (such as conditional autoregressive models) to a multivariate setup is proposed. Such models are applied to analyze the traffic crash data in the case of several types of events simulteneously. The Bayesian type approach is used. The main purpose of the work is to borrow strength or share information from similar sources, as well as to use the most directly avalible sources, to improve crash risk estimates. For this the several spatial priors for Bayesian multivariate hierarchial models and sufficient conditions to ensure posterior propriety under noninformative prior are provided.