previously all epochs were lowpass filtered at 40Hz ... although we decided to do this at 100Hz
fixed now ... but need to re-form epochs and re-run autoreject
is subj 07 still "bad" afterwards? are all others still "good"?
for time domain hypotheses: inspect data: if too noisy, need to filter at 40Hz in the respective script (amounts to two lowpass filters after one another, first 100Hz on raw, then 40Hz on epochs .. but is probably fine --> highpass filters are the more dangerous thing that may create artifacts when applied on epoched data, see also Luck 2014)
The PR contains an important fix: