I am afraid that this experiment is currently inactive.
Inactivating an experiment does delete all media files, as mentioned on the window that pops-up when a user is about changing the status of an experiment to “INACTIVE”. However, it appears that you downloaded those media files that had been saved (i.e. those not missing), on the 2nd of January.
Following up on Wake’s message, I do not believe that it is possible to retrieve the missing media files.
However I am currently looking at your experiment code to figure out what might have happened in the first place.
Looking at the code of the experiment, the issue is indeed very likely due to the 20s delay at the end of the experiment not being sufficient time for a long recording to be fully uploaded, especially if the participant’s network is not great, as you correctly mentioned @M11
Contrary to Wake’s comment, however, it is not the case that the audio files are uploaded at the end of the experiment: the upload starts at the end of the routine (in your case “InRoutineEnd”). However it is an asynchronous upload, i.e. the routine does not wait for the upload to be completed (since it might take too long).
A quick fix would then be to increase the 20s delay to something longer, although that would not garantee that all results would be saved under all circumstances.
A better fix would be to check whether the upload has completed, in In2Routine.
@M11 Will you be running this experiment again? If so, Wake and I could prepare a fix for you, that would ensure a proper saving of those audio files.
With my apologies for the delay in getting back to you, January is a rather crazy month.
Thank you very much Alain and wakecarter for your efforts and help! Loosing the data is really annoying because we tested experts you don’t often get. Yes, we plan another round of the experiment. It would be great if you could help me with a fix. What can I do?
We plan to collect more data for the above experiment by mid-February. If you could find the time to help me fix my problem that would be great, but I completely understand that you have limited resources. If time is too short to fix this problem, do you have any guideline times for how long a slow upload might take for a ~30 minute record? Maybe 40 seconds should be enough?
I’d recommend doing some testing since I feel like it’s an empirical question. It does seem likely that 40 seconds will be sufficient since 20 seconds was presumably enough in some cases (unless data was only saved when the participant left the tab open). You could also ask the participants to leave the tab open (though I’m not now sure if this will help).