E.g., participant sub-006 in the ERP CORE ERN dataset has 387 EEG epochs with matching triggers (i.e., () in epochs.drop_log) but the drop log also contains (NO_DATA) for the first event. So this event was not ignored when looking up the indices of the rejected epochs in get_bad_epochs, leading the bad_ixs to be off by 1.
Let's hope this is sufficiently general. Ideally, we should also take into account any pre-existing "bad" markers in the data, regardless of our own peak-to-peak rejection.
E.g., participant
sub-006in the ERP CORE ERN dataset has 387 EEG epochs with matching triggers (i.e.,()inepochs.drop_log) but the drop log also contains(NO_DATA)for the first event. So this event was not ignored when looking up the indices of the rejected epochs inget_bad_epochs, leading thebad_ixsto be off by 1.Let's hope this is sufficiently general. Ideally, we should also take into account any pre-existing "bad" markers in the data, regardless of our own peak-to-peak rejection.
Fixes #139