Spatial models in gretl: the SPM package

This article presents a new package for estimating cross-sectional spatial modelsingretl, named SPM. The package can handle three types of models: Spatial AutoregressiveModels (SAR), Spatial Durbin Models (SDM) and Spatial Error Models (SEM). The firstintegrates the canonical linear model by including spatial lags of the dependent variable, thesecond also includes spatial lags of independent variables, and the last examines spatiallyautoregressive errors. Computation of the Hessian matrix is performed in both analytical andmixed ways. Some speed-up procedures for the computation of the log-determinant termare implemented and compared. Finally, results of the proposed package are compared withthose of some software alternatives, namely Matlab Spatial Econometric Toolbox, Stata module sp and R packages spatialreg and spdep.