JATOS file export does not include metadata anymore
Study A - April/May 2020 study
OSWeb 1.3.3, JATOS 3.5.2 or 3.5.3
Procedure: Program Exp -> OSWeb -> Create JATOS file -> upload to JATOS -> Run study -> Download results through "Export results -> all"
The resulting JSON-coded txt file included JATOS variables like workerid, batchid or batchTitle.
Study B - October 2020 study
OSWeb 1.3.9, JATOS 3.5.7
Procedure: Program Exp -> OSWeb -> Create JATOS file -> upload to JATOS -> Run study -> Download results through "Export results -> all"
The resulting JSON-coded txt file does not include JATOS variables like workerid, batchid or batchTitle.
However, if I now download the results from the April/May study, these variables are still included in the original export file. I cannot seem to get the varibales in my files for the new study B. However, they are crucial as participant identifiers of course. What am I doing wrong?
Comments
So the thing is: JATOS has never automatically added any kind of metadata to the results. Any information about workerID, BatchID, etc that were these on Study A will have had to be added explicitly on your script, i.e. by OSWeb.
The reason for this is that JATOS doesn't know if your result data are in CSV, JSON or simple text format, so it cannot determine how to add the metadata information.
The fact that exporting your old results on the new JATOS works, shows that this is not a backend problem - the data are just not there on the results. I suspect that you changed something in the way you programmed your OSWeb study. I never worked with this directly - is there no checkbox to choose to include these variables? Or maybe something changed in OSWeb. Best is to tag @Daniel here and move this question to OSWeb.
hth
Elisa
Hi @marsaute,
This indeed changed as we wanted the OpenSesame data cleaned up. You can now download all this data from jatos by clicking `Export Metadata` next to the delete button in the Results page on Jatos. This way, the same data that does not change during the whole session, is not added to each logged row in the results file, effectively making them smaller.
hi @Daniel,
that is unfortunate. Well yes, I can download the metadata, but if I understand correctly, it has no variables to match it to the OSWeb data anymore. So I cannot investigate the metadata in conjunction with my study results. But that can be crucial! I need a participant identifier variable that is both in OSWeb data and JATOS metadata. Or am I missing something?
Now that you mention it, there is no variable shared by both sets through which they can be linked. We will need to reintroduce that.
I created an issue here: https://github.com/open-cogsci/opensesame-extension-osweb/issues/25
For your current situation, even though it is a very brittle assumption, I think the order of both sets should be the same. So If you take the 3rd row of metadata, it will correspond to the 3rd participant. It is of course impossible to verify this (hence the necessity to reintroduce a shared variable), but it may save the analysis of your current study.
Another option is that you parse the data exported by Jatos yourself (for instance through a script). Each participant result set in Jatos consists of two components:
The data stored in the context parameter is the metadata collected by Jatos.
Thanks @Daniel for planning to re-introduce it!
I guess I could check whether the orders of both data sets were the same in the April/May study in which I still have this identifier. It's not bullet-proof because both OSWeb and JATOS changed sicne then but it's something.
Regarding the parsing of the results data myself, are you sure that's possible? Maybe I misunderstood it, but the JSON results file I download from JATOS only includes the OSWeb data and no context parameter. So I'm not sure I get what the "partiipant result set in Jatos" is that you are referring to.
EDIT: There's a parameter with the name context but it only includes general browser-related information, no JATOS-specific stuff, except for the version number.
EDIT: There's a parameter with the name context but it only includes general browser-related information, no JATOS-specific stuff, except for the version number.
Right! I am sorry. It is a while ago I worked with it. Then an identifier definitely needs to be added again...
Thanks!