Development of a spatio-Temporal Autoregressive (STAR) Model Using Spatio-Temporal Weights Matrices

This paper addresses the development of a statistical model for spatial data collected over time, such as real estate data. A spatio-temporal autoregressive (STAR) model, based on spatial and temporal weight matrices, is proposed. The spatial and temporal weight matrices are used to develop simple spatio-temporal weight matrices. The model is obtained using existing spatio-temporal lag models (STLM) and spatial error models (SEM). The STAR model explicitly considers possible local temporal dynamic effects as well as spatial spillover effects given time reality. The model is then applied to empirical investigation using real estate data on apartments sold in Paris, between 1990 and 2001, and hedonic modelling using data.