Mariella and Tarantino propose the Spatial Temporal Conditional Auto-Regressive (STCAR) model for areal data, handling both spatial dependence between sites and temporal dependence across periods when a measurement is recorded at each location over a time interval. Inspired by the Generalized Multivariate CAR (GMCAR) model of Jin, Carlin & Banerjee (2005), the STCAR model reduces the unknown parameters to a single spatial-association parameter estimated at each period, and — unlike a Sims (1980) VAR — its space-time autoregressive matrix accounts for the spatial localization of the sampled realizations. The model is applied to reconstruct the spatial-temporal distribution of expected bankruptcies of small and medium enterprises in the province of Lecce, Italy.
"We propose a new model for areal data, the Spatial Temporal Conditional Auto-Regressive (STCAR) model, that allows to handle the spatial dependence between sites as well as the temporal dependence among the realizations … its space-time autoregressive matrix takes into account the spatial localization of the realizations sampled."
A tidy applied extension that bolts a time dimension onto the CAR machinery while keeping it parsimonious — one spatial-association parameter per period rather than a full space-time covariance. The framing that most repays attention is the contrast with a VAR: a Sims-style vector autoregression would let any area's past influence any other's future through an unrestricted coefficient matrix, whereas STCAR ties those cross-effects to geographic adjacency, trading generality for a spatial prior that is exactly right when influence really does travel through the map. The bankruptcy application is a cute demonstration that "areal disease risk" and "firm default risk" are the same modelling problem. It is a niche paper, but it usefully marks where the wiki's CAR material meets VAR dynamics.