Converting JATOS Results with Multiple Components to csv/xlsx
Hi y'all,
We have programmed a JATOS experiment that contains an OSWeb component, as well as 3 other components (i.e., consent form, pre-task questionnaire, and post-task questionnaire).
We are preparing to begin data collection and are planning our data preprocessing pipeline. We are using the "Convert JATOS results to csv/ xlsx" function in OSWeb. However, it does not seem to be converting the non-OSWeb components properly. I have attached a screenshot of the text file produced from JATOS for one test session, as well as the xlsx file produced by OSWeb. In the text file, each component is in a separate dictionary.
Please advise.
Thank you,
Bar
Comments
Hi @byos ,
I would start by deciding for yourself how you would like to organize the data. Basically, you have data from questionnaires and data from a trial-based experiment. These are fundamentally different forms of data, and so you have to decide on how you want to combine them so that you can analyze them easily. One option that comes to mind is:
Another option would be to have a column for each question from the questionnaires, and then repeat the response for all rows that belong to one participant. (Just like
subject_nr, say.)Once you've decided on a format, you can think of how to actually transform it. But first things first!
-- Sebastiaan
Check out SigmundAI.eu for our OpenSesame AI assistant!
Hi @byos,
If you're using different components collecting different types of data, the best option is to save the data in separate files using some information that allows you to cross-reference the data to know what data belong to what subject, and then recombine the data in the format that suits your analysis in Excel using functions like "match", "vlookup" etc., or some other program. The obvious piece of information would be the subject's id.
The alternative that Sebastiaan is highlighting can be very useful if your other components are fairly simple. It would consist in passing to the OS task the responses from the consent form and prequestionnaire and saving these on every trial by including it in the logger. Note that, obviously, the post-questionnaire data would not be part of task's output. Hence, if you're planning to condition the analysis of the data from the OS task on variables from the questionnaires, the first option (separate output files) is best, I think.
Not sure what your program your other components in, but I understand they all work under JATOS. I guess you can get generate a unique random subject id at the start and then pass it on to the different components in JATOS. Otherwise, an alternative that also work is to pass key information through the URL of your different components (some people use this method, for example, to pass info from Qualtrics to OSWeb and back, for example).
Hope this helps.
Fabrice.