Model Fitting: Robust Regression |
Method Tab |
You can use the Method tab to specify options for one of four robust regression algorithms:
The Method tab is shown in Figure 22.3. Each of the following UI controls corresponds to an option in the ROBUSTREG procedure.
specifies the algorithm used for the robust regression. The choices are M, LTS, S, and MM. This list corresponds to the METHOD= option in the PROC ROBUSTREG statement.
specifies the multiplier of the robust estimate of scale to use for outlier detection. This box corresponds to the CUTOFF= option in the MODEL statement.
specifies a cutoff value for leverage-point detection. This box corresponds to the CUTOFFALPHA= suboption of the LEVERAGE option in the MODEL statement.
The various methods each have options that are associated with them. When you select a method, the relevant options become active.
With METHOD=M, you can specify the following additional suboptions:
specifies a method for estimating the scale parameter. This list corresponds to the SCALE= option.
specifies the weight function used for the M estimate. This list corresponds to the WF= option.
With METHOD=LTS, you can specify the following additional suboptions:
specifies the intercept adjustment method in the LTS algorithm. Choosing "Default" corresponds to omitting the IADJUST= option. The other choices correspond to IADJUST=ALL or IADJUST=NONE.
With METHOD=S, you can specify the following additional suboptions:
specifies the choice of the function for the S estimator. This list corresponds to the CHIF= option.
specifies whether to refine for the S estimate. This check box corresponds to the NOREFINE option.
With METHOD=MM, you can specify the following additional suboptions:
specifies the initial estimator for the MM estimator. This list corresponds to the INITEST= option.
specifies the choice of the function for the MM estimator. This list corresponds to the CHIF= option.
specifies whether to display the bias test for the final MM estimate. This check box corresponds to the BIASTEST option.
Copyright © SAS Institute, Inc. All Rights Reserved.