(View the complete code for this example.)
You can fit more than one model by making only one call to PROC SPATIALREG. For example, if you want to fit both SAR and SEM models to the CRIMEOH data set, you can use the following statements:
proc spatialreg data=crimeoh Wmat=crimeWmat NONORMALIZE;
model crime=income hvalue / type=SAR;
model crime=income hvalue / type=SEM;
run;
The parameter estimates for the SAR and SEM models are shown in Output 31.3.1 and Output 31.3.2, respectively.
Output 31.3.1: Parameter Estimates of SAR Model
| Parameter Estimates | |||||
|---|---|---|---|---|---|
| Parameter | DF | Estimate | Standard Error |
t Value | Approx Pr > |t| |
| Intercept | 1 | 45.077070 | 7.870590 | 5.73 | <.0001 |
| income | 1 | -1.031531 | 0.328403 | -3.14 | 0.0017 |
| hvalue | 1 | -0.265924 | 0.088218 | -3.01 | 0.0026 |
| _rho | 1 | 0.431020 | 0.123594 | 3.49 | 0.0005 |
| _sigma2 | 1 | 95.487066 | 19.506312 | 4.90 | <.0001 |
Output 31.3.2: Parameter Estimates of SEM Model
| Parameter Estimates | |||||
|---|---|---|---|---|---|
| Parameter | DF | Estimate | Standard Error |
t Value | Approx Pr > |t| |
| Intercept | 1 | 59.891926 | 5.884117 | 10.18 | <.0001 |
| income | 1 | -0.941303 | 0.370268 | -2.54 | 0.0110 |
| hvalue | 1 | -0.302253 | 0.090552 | -3.34 | 0.0008 |
| _lambda | 1 | 0.561779 | 0.152413 | 3.69 | 0.0002 |
| _sigma2 | 1 | 95.572632 | 20.037633 | 4.77 | <.0001 |