SPATIALREG Procedure
Overview: SPATIALREG Procedure
The SPATIALREG (spatial regression) procedure analyzes spatial econometric models for cross-sectional data whose observations are spatially referenced or georeferenced. For example, housing price data that are collected from 48 continental states in the United States fall into the category of spatially referenced data. Compared to nonspatial regression models, spatial econometric models are capable of handling spatial interaction and spatial heterogeneity in a regression setting (Anselin 2001).
The SPATIALREG procedure supports the following models:
linear model
linear model with spatial lag of X (SLX) effects
spatial autoregressive (SAR) model
spatial Durbin model (SDM)
spatial error model (SEM)
spatial Durbin error model (SDEM)
spatial moving average (SMA) model
spatial Durbin moving average (SDMA) model
spatial autoregressive moving average (SARMA) model
spatial Durbin autoregressive moving average (SDARMA) model
spatial autoregressive confused (SAC) model
spatial Durbin autoregressive confused (SDAC) model
In general, SARMA, SDARMA, SAC, and SDAC models can require two spatial weights matrices. If you fit these four types of models by using the SPATIALREG procedure in SAS/ETS 14.2, the two spatial weights matrices are assumed to be identical.
Spatial econometric models have been widely used in economics, political science, sociology, and other fields. For example, LeSage and Pace (2009) provide a detailed introduction to commonly used spatial econometric models from both frequentist and Bayesian perspectives. A brief introduction to spatial econometric models is also provided by Elhorst (2013).
The SPATIALREG procedure in SAS/ETS 14.2 primarily uses the maximum likelihood estimation to achieve parameter estimation. Initial values for the nonlinear optimizations are usually calculated by ordinary least squares (OLS).