Usage Note 22641: Computing change in R-square and associated F- and p-values as variables are added to the regression model
In PROC REG, use the SCORR1(TESTS), SCORR1(SEQTESTS), or SCORR2(TESTS) options in the MODEL statement. Note that these options do not work with the model selection methods, the RIDGE or PCOMIT options, or with the GROUPNAMES= option.
These options produce F-tests which are calculated as the sum of squares for
the variable in question divided by the mean square error (MSE) for the full model.
If TESTS is specified, the MSE for the complete model (the MSE from the ANOVA
table) is used as the denominator. If SEQTESTS is specified, the full model
and its associated MSE change as variables are added sequentially to the
regression model.
For example, this statement:
model y=x1-x5/scorr1(seqtests);
produces the following table of parameter estimates:
Parameter Standard
Variable DF Estimate Error t Value Pr > |t|
Intercept 1 1.21230 0.18273 6.63 <.0001
X1 1 -0.15058 0.14669 -1.03 0.3059
X2 1 -0.05647 0.15469 -0.37 0.7155
X3 1 0.14848 0.14674 1.01 0.3129
X4 1 2.95669 0.13822 21.39 <.0001
X5 1 4.65109 0.15843 29.36 <.0001
Squared
Semi-partial Cumulative -Sequential Type I-
Variable DF Corr Type I R-Square F Value Pr > F
Intercept 1 . 0 . .
X1 1 0.00083606 0.00083606 0.17 0.6844
X2 1 0.00957 0.01041 1.91 0.1690
X3 1 0.00123 0.01163 0.24 0.6224
X4 1 0.33511 0.34675 100.03 <.0001
X5 1 0.53322 0.87997 861.81 <.0001
The column titled Squared Semi-partial Corr Type I contains the normal output from the SCORR1 optionthe partial R2 values. The Cumulative R-Square is the R2 from the previous model plus the partial R2 for the current variable. The final two columns give the sequential F-tests and their associated p-values.
Prior to SAS 8, PROC REG provides the partial R2 as variables are added to the model if you specify the SCORR1 option in the MODEL statement.
However, partial F-statistics and p-values are not available from the procedure.
Operating System and Release Information
*
For software releases that are not yet generally available, the Fixed
Release is the software release in which the problem is planned to be
fixed.
Type: | Usage Note |
Priority: | low |
Topic: | Analytics ==> Multivariate Analysis SAS Reference ==> Procedures ==> REG Analytics ==> Regression
|
Date Modified: | 2016-06-01 17:00:50 |
Date Created: | 2002-12-16 10:56:40 |