include customized question labels in the results file

[dgj2702]

When customizing scale labels under 'Customize questions', the customized labels do not appear correctly in the results. Instead the results file presents the globally defined labels. Note that as far as I can tell, the labels appear correctly to participants, just that the value that is recorded does not reflect the customized label.

For example, in a study asking participants to choose one of 7 article options in German (e.g., den, die, das, einen, eine, ein, [leer lassen]), some of the customized questions for the filler items have different forms (e.g., dem, der, das, einem, einer, ein, [leer lassen]). The participant sees the customized label and makes her choice -- for the sake of example, she chooses 'dem'. In the results file, however, the value appears as 'den', corresponding to the first option in the globally defined labels. I have double checked and the presentation appears correct and the result is easy enough to interpret as long as the customized labels are consistent, but it is not the expected behavior.

[gvarab]

Thank you for reporting this! Currently, this is the intended behavior; this ensures that the "summarized results" view works (i.e., that it possible and makes sense to count and summarize within a single column how many times e.g. response option nr. 1 was chosen). The idea is that what you customize is how the scale label appears for a specific item, but it is still mapped to one of the globally defined labels internally for the purpose of saving and displaying the results.

How I go about using customized scales: in the global setting, I define abstract labels that are interpretable for me in the results (e.g. "def-m,def-f,def-n,indef-m,..."), and I overwrite that for each item in my custom csv file – then the chosen options will be aggregated correctly even if "def-m" sometimes appears as "den" and sometimes as "dem".

But I definitely see that this is not clear from the current interface – it might be clearer if there was an explicit choice between global setting / custom setting, and in the latter case, the saved variable would be an abstract placeholder like "option1,option2...". We decided to allow mixing both options because sometimes it's useful to "initialize" all items with a global scale and then only adjust some of them individually (like you did, I think). I'll try to re-think how the interface could be clearer in this respect, or if there's generally a better solution.

(related issue, where I got confused myself about the same thing: https://github.com/2e2a/l-rex/issues/101)

[dgj2702]

How I go about using customized scales: in the global setting, I define abstract labels that are interpretable for me in the results (e.g. "def-m,def-f,def-n,indef-m,..."), and I overwrite that for each item in my custom csv file – then the chosen options will be aggregated correctly even if "def-m" sometimes appears as "den" and sometimes as "dem".

That makes complete sense and would in fact be preferable to my approach (i.e., adjusting labels individually). Perhaps the interface can be split into "Scale values" and "Scale labels", something like:

Scale values Rating scale values, separated by commas (e.g. "1,2,3,4,5"). If a label contains a comma itself, escape it with "\" (e.g. "A,B,Can't decide\, I like both"). These values will be recorded in the results.

Scale labels (global) How the scale values appear to participants. This can be left blank if the scale labels should be identical to the scale values defined above. Note that scale labels can also be defined in the materials, either individually per item or in the uploaded csv file. The latter is especially useful when labels differ across items.

Admittedly the above suggestion is rather complicated for new users, but maybe it helps for brainstorming some ideas.

[gvarab]

I like your suggestion! A slightly less flexible but maybe easier/clearer version of it would be to always use something abstract like "1,2,3,4,5" as the scale values, so that the user would only have to set the number of options and (optionally) the scale labels (global and/or item-specific). Then the step of mapping, e.g., 1 to def-m, 2 to def-f, etc., would not happen within L-Rex but would be up to the researcher when interpreting/handling the result file.

[dgj2702]

Yes, that would also work.