[Eeglablist] Re-referencing drastically changes my power spectrum

[Ulrich Pomper]

Dear Seana,

Many thanks for your reply! Yes, that makes sense, and I also agree that 
re-referencing first is probably the safest option.
However, I try to re-refernce only after interpolating noisy channels, 
in order to not introduce that noise to other channels. And in my 
experience, it is more convenient to identify noisy channels after the 
data are filtered and epoched.

Regardless, I'm still puzzled that performing those two preprocessing 
steps in two different orders produces so radically different results, 
with one clearly looking wrong.

Kind regards,

Ulrich

Am 18.08.2024 18:39, schrieb Seana Coulson:

> Well, first I was thinking that re-referencing is linear so it 
> shouldn't matter when you do it. But upon further reflection, if you 
> re-reference after filtering, the quantity subtracted from every 
> channel will differ for the low frequency data and the high frequency 
> data which would result in differences in the overall activity. That 
> is, band pass filtering is not a linear operation so it does matter 
> when you do it.
> In general, it's customary to do re-referencing before other data 
> processing steps (and I think you've discovered why for yourself!). In 
> your case, it would definitely be best for you to first compute the 
> average reference data and then do the band pass filtering for your 
> analysis.
> Seana
> 
> On Sun, Aug 18, 2024 at 4:50 AM Ulrich Pomper via eeglablist 
> <eeglablist at sccn.ucsd.edu> wrote:
> 
>> Dear EEGlab list,
>> 
>> I have a question regarding to order of two common preprocessing 
>> steps:
>> filtering and re-referencing to common average.
>> 
>> In my current EEG study, I initially filter the raw, continuous data
>> into low (2-12 Hz) and high frequency components (13 - 80 Hz), and
>> afterwards re-reference the data to common average. However, this
>> re-referencing step really messes up the spectrum of the data in a way
>> that makes me suspect an error (please see panels 7 and 8 in this 
>> figure
>> [1]).
>> 
>> When I re-reference first and filter second, on the other hand, the
>> spectrum stays intact and the data seem OK (see panels 5 and 6).
>> 
>> I was hoping that someone might be able to explain to me what is going
>> on here, and potentially confirm that re-referencing first and 
>> filtering
>> second is actually the correct way to go.
>> 
>> I'd highly appreciate any input!
>> 
>> Many thanks,
>> 
>> Ulrich
>> 
>> --
>> Dr. Ulrich Pomper
>> 
>> Senior Scientist
>> Department of Cognition, Emotion, and Methods in Psychology
>> University of Vienna
>> Liebiggasse 5
>> 1010 Vienna
>> Austria
>> 
>> https://urldefense.com/v3/__http://www.rhythmsincognition.org__;!!Mih3wA!FGS_Heq7qxqfOnJQRY9iWkPCeyqHn2JXpN813UzKIfwO9NvF21ZtZ0Kice2TBxBkbfGiZ_gKEek0myyUErlLBlsFWQu_mv1bog$ 
>>  [2]
>> ulrich.pomper at univie.ac.at
>> +43-1-4277-47149 [3]
>> 
>> Links:
>> ------
>> [1]
>> https://urldefense.com/v3/__https://drive.google.com/file/d/1iBnYCQiLmWAEd1rLMLcHYqD3AHbD5dRP/view?usp=sharing__;!!Mih3wA!FGS_Heq7qxqfOnJQRY9iWkPCeyqHn2JXpN813UzKIfwO9NvF21ZtZ0Kice2TBxBkbfGiZ_gKEek0myyUErlLBlsFWQt8C5gzOA$
>> [2] 
>> https://urldefense.com/v3/__http://www.rhythmsincognition.org__;!!Mih3wA!FGS_Heq7qxqfOnJQRY9iWkPCeyqHn2JXpN813UzKIfwO9NvF21ZtZ0Kice2TBxBkbfGiZ_gKEek0myyUErlLBlsFWQu_mv1bog$
>> [3] tel:+431-4277-47149
>> _______________________________________________
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>> To unsubscribe, send an empty email to 
>> eeglablist-unsubscribe at sccn.ucsd.edu
>> For digest mode, send an email with the subject "set digest mime" to 
>> eeglablist-request at sccn.ucsd.edu