URL of experiment: https://gitlab.pavlovia.org/petyaraczbme/lendulet_bme_szokiserlet_vegso/html
Description of the problem:
Process: Originally, I enabled partial data collection. I export data from pavlovia using the web interface.
Expected outcome: When I export data from pavlovia, I expect to see a large csv with sessions submitted successfully, as well as logs and csv-s for incomplete sessions. This works for my other experiments.
Actual outcome: There are no csv-s for incomplete sessions in the data drop.
I think I’ve read somewhere that migration after the data centre fire affected data in such a way that pavlovia decided to prioritise recovering complete rather than partial data. But I can’t find that blog post anymore.
I’m quite sure this is not related with the data centre fire, but it could be related with the way participants quit your experiment.
- If they quit it by pressing the escape key, partial data is submitted to the server.
- If they quit it by closing the browser window, PsychoJS will still make an attempt to send the data before the program is aborted. However, browser have been making it harder to do so, as such a feature (have something happen if a window closes) can be abused by less scrupulous software.
I took a closer look. Another factor could be the amount of data already collected; the technique we use is intended for relatively small amounts, so over a certain threshold they could malfunction. I’ll pitch with the devteam if we can’t improve on this perhaps.