search

ml4ai / tomcat

ToMCAT: Theory of Mind-based Cognitive Architecture for Teams
https://ml4ai.github.io/tomcat/
MIT License
21 stars 7 forks source link

Make eeg data for experiment `exp_2022_10_24_12` invalid #539

Open paulosaurusrex opened 1 year ago

paulosaurusrex commented 1 year ago

In the previous implementation, EEG signals for experiment exp_2022_10_24_12 were ignored and not saved to the database because Eric detected some issues with the timestamps of the signals. EEG data for this experiment was added to the new dataset implementation (using Postgres). At first glance, timestamps look normal, so we need further detail about the problem. We can later either remove all EEG data from this experiment from the eeg_raw table or use the data_validity to mark portions of it as invalid.

@eduongAZ, can you give us more detail about the issue with EEG data in this trial? I noticed the timestamps seem to be decreasing instead of increasing in time. Is that the problem?

eduongAZ commented 1 year ago

exp_2022_10_24_12 for EEG is strange because because we had a few crashes during the experiment:

Note from @rchamplin

00023 on Lion.
00027 was a no show, EA (99905) sat in on Tiger with no cap on.
00051 on Leopard.
For some unknown reason, actiChamps app froze up on EEG computer during HOT causing EEG streams to stop. This happened two times. After first time, I restarted actiChamps instances and the EEG's streamed for a few minutes. After failing a second time, I disconnected EEG battery packs, rebooted EEG computer, reconnected batteries, restarted actiChamps instances, restarted streaming. EEG's streamed ok for rest of experiment.
During Saturn A, participant on Leopard started feeling ill. She said she needed to go to restroom. EA disconnected her cap. This happened with about 2.5 minutes left in Saturn A. She was not able to continue. EA decided to stop experiment and all recording.

Initially the timestamps are fine, but then it starts counting backward, all the way to negative seconds, which is very strange.

paulosaurusrex commented 1 year ago

Thanks for the explanation. I'm wondering if it's worth leaving the eeg data for this experiment in the table of if we should just remove it altogether. @adarshp, do you have a preference?

eduongAZ commented 1 year ago

@paulosoaresua I think we might be able to repair some data. We should look for the timestamps that we expect to be correct. When I first parse the file, the initial timestamps are correct. I think we can cut off the timestamps we suspect to be caused by crashing actiCHamps.

adarshp commented 1 year ago

I think we should remove part of the data --- so far, I've been thinking of data within a task for a particular group session ID and modality as 'atomic', as in, either the data is good, or it is not. So, I think it's worth keeping the EEG data for this group session for all the tasks/phases for which there are no incorrect timestamps (i.e., we should use the data_validity table to mark portions of the data as invalid)