URL of experiment: : https://run.pavlovia.org/cherkasovabbalab/reinforcement_learning_bandit
Description of the problem: For one of my participants, there are some data missing from the data file - even though all the trials are recorded. The experiment collects mouse click responses. The file has the data indicating whether the response was correct or not, but is missing data (empty brackets) on the position of the mouse click, the identity of the stimulus clicked, and the response times. This is true for most of the trials following the initial few. I was wondering what might be causing this. And also whether this might be related to issues reported by a couple of users, i.e. task being ‘glitchy’ and not accepting their responses (causing them to quit the task and hence no data). I have not been able to reproduce these issues on my end. Even when I temporarily disconnect my network, this does not cause the task to be glitchy. These do not appear to be browser-related issues: we instruct participants to use Chrome or Firefox, and participants who had these issues were on Chrome. Any suggestions as to what may be causing this or how to debug this would be greatly appreciated. E.g. is it possible this would be caused by firewalls, adblockers, etc?
Thanks for the heads-up! Could you ask which operating system they were using? (Mac OS X perhaps?) And which one do you use? (Windows 10 perhaps?)
About network issues: if the network connection drops, this shouldn’t interfere with mouse responses, though it does mean the data won’t be sent to the servers.
Hello Thomas, thank you for your response.
One of the participants was using Mac OS and the rest were using Windows. We use Windows 10 on our end. Does the operating system make a difference? Should we be restricting OS as well as browsers for study entry?
There was another issue with a participant who was using Chrome and Windows 10. She reported that the task was ‘lagging’. By lagging she meant a screen lag and a mouse freeze. For example, she would select a stimulus, but there would be a lag in response, so she would perform another mouse click to select the same stimulus again. On the next trial, the task would sometimes select the stimulus for her, presumably because of the multiple mouse clicks being registered. (Of course, I cannot use her data.) What do you think could be causing this?
Safari on MacOS has a tendency to misbehave, but Chrome should be all right. Could you share your gitlab so that I can take a peek at your experiment? Searching for experiments on Pavlovia — PsychoPy v2021.1
Yes, of course, I invited you as guest to the project (hopefully, the right Thomas Pronk). Or do you need developer permissions?
I normally use Windows 10 and Chrome and have tested the task multiple times - never encountered any issues on my end…
No prob, but I’m afraid I don’t have access yet. I’ll need to be developer. My gitlab username is tpronk btw
Hi Thomas, I granted you permissions as developer. I added the wrong Thomas Pronk before… Thanks!
I’m in! I’ll take a peek next week
Hey again! I examined your experiment a bit and I’ve got two recommendations:
- The experiment exports to PsychoJS 2020.2.10; consider upgrading to 2021.1.4
- I see the Mouse component saves state on every frame; consider changing it to “on click” and have the code that sets up feedback run in and “end routine” tab instead of “each frame”