Is it possible to add groups to a form (to incorporate the field-list appearance) after data collection has started, without messing up the submitted data?
That is, some way to group questions using field-list that avoids this warning:
Workflow warnings:
- The following fields have been deleted, renamed or are now in different groups or repeats. These fields will not be visible in the Submission table or included in exports by default. Learn more.
Fields: ...
This might not be the most elegant solution, but it does the job. You may try something similar:
Initially, I designed the XLSForm like this:
| type | name | label | appearance | calculation |
|---|---|---|---|---|
| int | some_integer | Enter a number | ||
| text | some_text | Enter some text | ||
| select_one list1 | some_chooseone | Choose one |
Later on, I wanted these three questions / variables to appear together on a single screen, so I grouped them as new variables and changed the original ones into calculate types that simply mirror the values from within the group:
| type | name | label | appearance | calculation |
|---|---|---|---|---|
| calculate | some_integer | Enter a number | ${some_integer2} | |
| calculate | some_text | Enter some text | ${some_text2} | |
| calculate | some_chooseone | Choose one | ${some_chooseone2} | |
| begin_group | group1 | New Group | field-list | |
| int | some_integer2 | Enter a number | ||
| text | some_text2 | Enter some text | ||
| select_one list1 | some_chooseone2 | Choose one | ||
| end_group |
Yes, this approach adds three (or more in your case) extra fields to the form, which introduces a bit of redundancy. But it helps with data analysis, since the original variables still hold the newly added data - keeping data consistent and making downstream work easier.
A problem might be that with the change to a calculate type the previous data could no more be changed/edited (manually), except by temporarily switching back the type again in a current form version.
There are also negative effects if the type switch adresses a previously select type, as e.g. SPSS labels will get lost now for export.
Most if the time, we try hard to avoid such changes during data collection through systematic pretesting. Furthermore, we prefer to adapt data structure issues externally, after export.
