The Start Groups node in SAS^® Enterprise Miner™ enables you to specify Cross-Validation as the Mode. (The processing uses the levels of one or more Stratification variables that you specify.) However, when the corresponding End Groups node generates Scoring SAS code, the code is not correct. The assessment of the final results is incorrect because the assessment is based on the scored table. If you score new data using the Scoring SAS code, then the scores are not correct.

There are no errors or warnings to indicate a problem.

If you are applying the score code manually , then you can manually edit the score code to correct it before applying it. Applying score code manually involves using a DATA step to perform the scoring, instead of using a Score node. To correct the code, edit these lines in the score code. The number of lines depends on the number of levels of Stratification variable(s) in your data.

In each of those lines, remove the "not" symbol (^) so that the edited line looks like this:

EXAMPLE

The sample data set SAMPSIO.DMEXA1 is used to create a flow. In the Start Groups node, variables EDLEVEL and GENDER are specified as Stratification variables. The Mode is set to Cross-Validation.

After the flow runs, the End Groups node Scoring SAS code contains these lines:

...and so on, through the possible combinations of _FORMAT1_11 and GENDER. Save the Scoring SAS code from the Score node. Edit the saved code. In each of those lines, remove the "not" symbol (^) so that the edited lines look like these:

Operating System and Release Information