[Eeglablist] Question about Order of rereferencing and artifact removal

[Arnaud Delorme]

Hi Steven,

About ICA, it is usually better to re-reference before running ICA although it is also possible to re-reference after running ICA (in which case the ICA matrices are re-referenced as well).
The choice of reference does not generally affect the ICA solution dramatically although, sometimes, I think average reference is performing better (this is a judgement that is by no mean based on rigorous comparisons). It is critical to remove bad channels before running ICA though and before referencing the data (especially average reference which will use all channels to compute the new reference). 

Best,

Arno

On 22 Mar 2013, at 03:52, Stephen Politzer-Ahles wrote:

> Hi Steven,
> 
> As for the order of operations, in our lab we always re-reference first, but I don't in principle see why doing artifact rejection first would affect the referencing process--given that at each datapoint you're just subtracting a value from all channels, that process shouldn't be affected by whether or not you removed other data points. The only way I can imagine it would have an effect is that, when you do artifact rejection, you would have to keep in mind that artifacts might 'look' different depending on how the data are referenced (for instance, blinks might show a different polarity, etc.). I've only done re-referencing to mastoids, not to average reference, but according to my understanding the concept should be the same.
> 
> I'm not sure about the ICA issue, let's see what the others on the list have to say...
> 
> Best,
> Steve
> 
> On Thu, Mar 21, 2013 at 5:10 PM, Steven Pillen <stevendpillen at gmail.com> wrote:
> Hello, EEGLABList.
> 
> There was a question we had at our lab about the order of operations when it comes to rereferencing and other steps of preprocessing.  Is considered better to remove artifactual data before rereferencing to the average, or before filtering the data? 
> 
> In addition, when we tried to rereference to the average immediately after removing artifactual data, when we got to the point where we tried to run ICA, we got an error message that read:
> 
> "EEGLAB has detectat that the rank of your data matrix is lower the number [sic] of input data channels.  This might be because you are including a reference channel or because you are running a second ICA decomposition.  The proposed dimension for ICA is 31 (out of 32 channels). Rank computation may be innaccurate so you may edit this number below.  If you do not understand, simply press OK."
> 
> It would not run with either 31, or 32 channels, and we're not sure why.
> 
> Does anyone have any insight as to why?
> 
> Thank you,
> Steven Pillen
> 
> _______________________________________________
> 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
> 
> 
> 
> -- 
> Stephen Politzer-Ahles
> University of Kansas
> Linguistics Department
> http://people.ku.edu/~sjpa/ _______________________________________________
> 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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130328/c67d6ef7/attachment.html>