IMPROVED ESTIMATION AND PREDICTION IN SPATIAL MODELS

Abstract

In recent years, there has been a growing interest in specification and newlinedeveloping inference procedures for spatial econometric models. These models newlinehave found enormous applications in various fields such as geography, newlineeconomics, geosciences, demography etc. Bustos et al (2009) applied spatial newlineARMA models for image filtering. For modeling causal relationships for newlinespatially referenced data keeping in view the presence of spatial dependence in newlineobservations, these models either incorporate errors having spatial newlineautocorrelation (spatial error model) or by including dependent variable having newlinespatial autocorrelation (spatial lag model). The spatial weight matrix with newlineknown weights represents a priori understanding of the nature of spatial newlineinterdependence between different geographical regions or between different newlineeconomic agents. For theoretical overviews of spatial econometrics one may newlinerefer to Anselin (1988). Le Sage and Pace (2009) provides introduction to newlinespatial econometric modeling along with its various applications and discusses newlineclassical and Bayesian inference procedures for spatial autoregressive (SAR) newlinemodel, spatial Durbin model (SDM), and spatial error model (SEM). newline