Hi,

We're upgrading from 1.85 to 1.90+ Build 9118. I found a difference in the export of datafiles (SPSS and R).
in LS 1.85 the -oth- values in the database (from the "Other" option in eg. a List type question), were stripped so that they appear as a missing value in the exported data file (.dat or .csv).
In 1.90 Now the -oth- value is in the exported data file. Since the type this field is numeric it gives warning messages wen importing the file in R. SPSS does not seem the give warnings, but still it is bit strange to have character values in a numeric field. We use this export also for an export to SAS and there it als generates warnings.

I think the stripping of the "-oth-" text was done in export_data_functions.php with the function strip_tags_full. The line that converts the -oth- values has been commented in LS 1.9.
Do you know what the reason was ? Should it have been handled in the new FlattenText function ?

Thanks for any feedback.
Jan.

And after thinking a bit more...

I bumped on this with questions that had only numeric answercodes.
But it actually makes sense not to filter the -oth- in case the answer codes are character codes. then you can keep -oth- in the data and it is clearly different from a missing value. So you don't want to lose that information.

So -oth- should only be filtered when you have numeric answercodes, or it should be changed to a special value eg. 99999, but that's dangerous as wel. But I guess that this is not an upgrade issue anymore.

We'll advise users to use character answercodes and explain why.

Cheers,
Jan.

If you think it's a bug or it can be easily imporved, please open a ticket at our bugtracker.