Output from POWER UNIVARIATE LINEAR

Posted by Bruce Weaver on
URL: http://spssx-discussion.165.s1.nabble.com/Output-from-POWER-UNIVARIATE-LINEAR-tp5740814.html

I just noticed that SPSS 27 added some POWER commands.  I decided to try out POWER UNIVARIATE LINEAR on the first example shown on this Stata page:

https://www.stata.com/features/overview/power-analysis-for-linear-regression-models/

Scroll down to "Here, we demonstrate PSS for an R2 test of a subset of coefficients in a multiple linear regression" to find the example.  The output shows that n = 81 is needed to ensure power of 0.8 to detect a change in R^2 of 0.1 (from 0.1 to 0.2).  

I used the following SPSS code to carry out the same analysis:

POWER UNIVARIATE LINEAR
  /PARAMETERS MODEL=FIXED SIGNIFICANCE=0.05 POWER=0.8
   TOTAL_PREDICTORS=5 TEST_PREDICTORS=2
    FULL_MODEL=0.2 NESTED_MODEL=0.1 INTERCEPT =TRUE.


The results can be seen in this image:



SPSS gives the same estimated sample size as Stata:  n = 81.  That is not the problem.  

Here is the problem:  The SPSS output shows there are 5 predictors in the full model, and 2 predictors in the "Nested" model.  That labeling makes no sense to me given this excerpt from the command syntax:

   TOTAL_PREDICTORS=5 TEST_PREDICTORS=2

If there are 5 variables in the full model, and 2 predictors that are being tested (for their impact on the change in Rsq), then surely there are 5-2 = 3 variables in the nested, or reduced model.  (In the Stata output, this is shown as ncontrol = 3 and ntested = 2.)  

How are others interpreting the labeling of the Full and Nested columns under Predictors?  Is there a mistake in the labeling?  Or am I misunderstanding something obvious?

Cheers,
Bruce