Bayesian inference in STAR models using neighbourhood effects

Abstract

In this paper we propose a semi-parametric Bayesian analysis of a spatio-temporal autoregressive model (STAR) with neighbourhood effects similar to those of Pace et al. (1998, 2000). This approach allows us to make inferences about the parameters of the model and, more particularly, about the number of neighbours, without having to appeal to asymptotic results. In addition, the procedure used to obtain the out-sample predictions takes into account the uncertainty associated to the estimation of the model parameters in a more realistic way. The methodology is illustrated by means of an application to the real estate market in the city of Zaragoza (Spain).