There, not only did we remove bad epochs, but also interpolated data before submission to ICA fitting. However, this is not what is recommended in the autoreject documentation: there, the suggestion is to only remove the bad epochs, not interpolate.
This PR fixes the behavior accordingly and updates the documentation and log messages.
I tested on a set of noise EEG data and I'm actually getting better results with this approach for the majority participants .
Before merging …
[x] Changelog has been updated (docs/source/changes.md)
This is a fixup for #810
There, not only did we remove bad epochs, but also interpolated data before submission to ICA fitting. However, this is not what is recommended in the autoreject documentation: there, the suggestion is to only remove the bad epochs, not interpolate.
This PR fixes the behavior accordingly and updates the documentation and log messages.
I tested on a set of noise EEG data and I'm actually getting better results with this approach for the majority participants .
Before merging …
docs/source/changes.md
)