PLM restore not scoring because of variable format

[Reeza]

Can you confirm the variable type in each data set?
is STAT_MAKE a character in all data sets? It seems to be treating it as a character variable. Unfortunately type issues are common when importing from Excel files.

[Reeza]

Is there more to the warning. The link (SAS Notes) indicates this is an issue with continuous variables, which are typically numeric and BEST is also numeric so that shouldn't apply to a character variable. Also, the error screenshotted in the post has more information than yours shown, for one it shows both the expected and actual formats.

[Negarev]

No, this is pretty much it... It doesn't specify the expected format

[STAT_Kathleen]

The problem associated SAS Problem Note:

   http://support.sas.com/kb/42/902.html

has been fixed in SAS 9.3.

I suspect there is a different reason that your SAS Log reports the messages:

  WARNING: The '$' format for the variable XYZ in the secondary data set is not compatible with the    format currently applied ('$')…..

Based upon the warning it seems that either the formats differ or there are different variables in the data sets.

I would suggest doing a PROC CONTENTS on the data set that is being used in the SAS/STAT procedure that was used to create the item store and the data set in the SCORE statement in the PROC PLM step to verify that both data sets have the exactly the same variables , same formats etc.  Please let me know if the results of using PROC CONTENTS on the two data sets.

If the results of the PROC CONTENTS are identical then you will need you to open a track and contact Technical Support.  You will need to provide the input data set , the PROC FORMAT and SAS/STAT code that created the item store and PLM code; then I will be happy to test your code to provide further recommendations.

[Tom]

---

@Negarev wrote:

No, this is pretty much it... It doesn't specify the expected format

---

So it looks like the issue with the incorrect format is just a warning, your step should stilll run.

But the second one where your dataset does not have all of the variables used in the model is a bigger problem.

Use the SHOW ALL command to see what variables are used in the model stored in the item store and make sure your datasets have all of the variables.