The CSPATIALREG Procedure

Example in this section uses the CrimeOH and CrimeWmat data sets that are described in Columbus Crime Data to demonstrate fitting and comparison of multiple models in PROC CSPATIALREG.

The following statements fit multiple models to the CrimeOH and CrimeWmat data sets, which are described in Columbus Crime Data, using the regressors Income and HValue:

/*-- TYPE=AUTO --*/
proc cspatialreg data=mycas.CrimeOH Wmat=mycas.CrimeWmat;
   model crime=income hvalue / type=AUTO;
   spatialid sid;
run;

In this example, the TYPE=AUTO option is specified in the MODEL statement to fit seven models—SAC, SARMA, SAR, SEM, SMA, CAR, and LINEAR. However, the CAR model is not fitted because the input spatial weights matrix is not symmetric. The parameter estimates for the six remaining models—SAC, SARMA, SAR, SEM, SMA, and LINEAR—are shown in Output 12.6.1. The comparison of these six models is provided in Output 12.6.2. If AIC is used for model comparison, Output 12.6.2 suggests that the SAR model is the best model among the six candidate models, because its AIC value is the smallest.