Problem Note 15214: Incorrect predictions are possible for the additional data data set
In the Predictions page of the task, if you mark the "Additional data"
checkbox to create predictions for (i.e. "score") a separate data set,
the task incorrectly uses the observations in the additional data to fit
the model. This will cause problems if there are observations in the
"additional data" data set where all variables used in the analysis are
nonmissing. The additional data to be scored should not affect the
model fit.
If you do not have a Frequency Count variable designated in the Task
Roles page of the task, you can correct the problem as follows. Mark
the "Preview code" checkbox at the bottom of the opened task. This
displays a window containing the SAS code that the task will submit. In
the code window, click the "Insert code" button at the top. In the
"User Code" window that opens, double-click on the "<double-click to
insert code>" line that appears immediately before the word "MODEL"
which begins the MODEL statement. In the "Enter user code" window that
opens, enter this line:
FREQ _FREQ_;
then click OK. Click OK again to close the "User Code" window. You may
continue to work with the open task window. When you run the task, the
added statement will ensure that the data to be scored do not affect the
model fit, but are correctly scored.
Select the Hot Fix tab in this note to access the hot fix for this issue.
Operating System and Release Information
| SAS System | SAS Enterprise Guide | Microsoft Windows XP Professional | 2.1 | 4.1 | 9.1 TS1M2 | 9.1 TS1M3 SP1 |
| Microsoft Windows NT Workstation | 2.1 | 4.1 | 9.1 TS1M2 | 9.1 TS1M3 SP1 |
| Microsoft Windows 2000 Professional | 2.1 | 4.1 | 9.1 TS1M2 | 9.1 TS1M3 SP1 |
*
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: | Problem Note |
| Priority: | alert |
| Date Modified: | 2005-07-15 10:09:32 |
| Date Created: | 2005-05-16 11:01:54 |
