The Method Tab
You can use the Method tab to specify options for the loess algorithm.
The following options are available:
- Selection method
-
specifies how to choose the loess smoothing parameter. This option
corresponds to the SELECT= option in the MODEL statement.
- AICC
-
selects the smoothing parameter that minimizes the corrected Akaike information
criterion.
- GCV
-
selects the smoothing parameter that minimizes the generalized
cross validation criterion.
- Approx. model DF
-
selects the smoothing parameter for which the trace of the prediction
matrix is closest to the Target model DF. This corresponds to the
SELECT=DF1 option in the MODEL statement.
- Manual
-
enables you to specify the value in the Smoothing parameter field.
- Exhaustive search for minimum
-
specifies that a global minimum be found within the range of smoothing
parameter values examined. This corresponds to the GLOBAL modifier to
the SELECT= option in the MODEL statement.
This option is computationally expensive.
- Restrict search range
-
specifies that only smoothing parameters greater than or equal to
Lower bound and less than or equal to Upper bound be examined.
- Robust reweighting iterations
-
specifies the number of iterative reweighting steps.
Stat Studio counts the initial fit as the th
reweighting iteration. This differs from the LOESS procedure,
which counts the initial fit as the first
iteration. Thus if you type in this field,
the option corresponds to ITERATIONS= in the MODEL statement.
- Interpolating polynomial
-
specifies whether the interpolating polynomial is linear or cubic.
This corresponds
to the INTERP= option in the MODEL statement.
- Calculation of lookup DF
-
specifies the method used to calculate the "lookup" degrees of
freedom used in performing statistical inference. This corresponds to
the DFMETHOD= option in the MODEL statement.
- Local regression polynomial
-
specifies the degree of the local polynomial to use for each local
regression. The choice is linear or quadratic.
This corresponds to the DEGREE= option in the MODEL statement.
Copyright © 2008 by SAS Institute Inc., Cary, NC, USA. All rights reserved.