Markup language for questionnaires

[Stian Håklev]

In my work with educational research, I often encounter questionnaires. I wonder if anyone know about existing or propose markup-languages for writing questionnaires? I did a bit of searching, but found only some academic papers and some mentions of heavy XML frameworks from the 1990's - nothing that looked very current or useful.

My eventual vision would be to have something very light-weight, perhaps Markdown-based, and be able to generate both web and paper questionnaires. (I don't know if there's any good questionnaire websites that have APIs for ingesting or importing question setups?)... 

The second step would be able to automatically generate some R code, for example, to parse the incoming data from the web questionnaire service. I like Google Forms, but I always end up having to write a bunch of boiler plate, to change field names, convert things to ordered factors, etc. I should be able to specify in the questionnaire markup file that something is an ordered factor (like a Likert-type Not at all, somewhat, neutral etc), and get the data cleanup for free...

Another advantage of a simple text-based format is that it would make it much easier to share and fork, diff etc questionnaires. There is a huge amount of standard questionnaires in educational science, for example, but many of them are encumbered with Copyright and high fee payments. And even if they are not, you are likely just to get them as PDF and having to retype them. I'd love a Github repo full of open sourced validated instruments for testing for example physics knowledge, and being able to fork one these, make a few changes, and right away see what has been changed, generate paper and online questionnaires, automatically clean up the data etc...

So yeah, lot's of ideas. I'd love to hear if anyone else finds this interesting, or if you could point me in the direction of people who are already working on this kind of stuff.

Stian

--

http://reganmian.net/blog -- Random Stuff that Matters

[Thomas Leeper]

There are two standards worth looking at here:

1) Data Documentation Initiative: http://www.ddialliance.org/ This is the current international standard for survey preservation, which includes both survey data and metadata. I would say it is borderline impossible to generate by hand. I've had the ambition to create an R package to generate it, but haven't had the time. There are a couple of R packages floating around for generating DDI from SPSS (http://www.icesi.edu.co/CRAN/web/packages/spssDDI/index.html) and Stata (https://github.com/ddionrails/r2ddi) files, but nothing to generate them from scratch or really do anything else with them.

2) Triple-S: http://www.triple-s.org/. This is probably the 90's era standard you've encountered. It's only sophisticated enough to handle relatively basic surveys, but seems to be widely supported by a lot of software. You might be able to write this by hand and/or craft a simple package to do it for you.

Additionally, if you use Qualtrics, you should take a look at the features for survey import, which include a decent markdown-esque markup language (http://www.qualtrics.com/university/researchsuite/advanced-building/advanced-options-drop-down/import-and-export-surveys/#PreparingATXTFileInAdvancedFormatForImporting). Perhaps that would be the model for a more general survey markup language that could be automatically converted to DDI, Triple-S, or whatever other format.

This is definitely something I've thought about working on, especially the idea of generating R code to process survey results based on a questionnaire markup document. But I want to try to avoid the XKCD standards problem (http://xkcd.com/927/). So, if you have ideas for how to make this work, I'd love to discuss more.