[Eeglablist] Artifacts after removal of ICA component

[Iman Mohammad-Rezazadeh]

You may want to see https://sccn.ucsd.edu/pipermail/eeglablist/2014/007810.html
-Iman

-------------------------------------------------------------
Iman M.Rezazadeh, Ph.D
Semel Intitute, UCLA , Los Angeles
& Center for Mind and Brain, UC DAVIS, Davis



From: eeglablist-bounces at sccn.ucsd.edu [mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Ahmad, Jumana
Sent: Friday, December 9, 2016 10:02 AM
To: Eric HG <erichg2013 at gmail.com>; eeglablist <eeglablist at sccn.ucsd.edu>
Subject: Re: [Eeglablist] Artifacts after removal of ICA component

Hi Eric,

Did you reduce the rank of your data by 1 before running ICA? I would do Matlab rank(EEG.data(:,:)) and check it is not 1 less than the number of your channels in ICA. If it is the same and you are not rank deficient. If you are then you can perform channel rejection to match EEG.nbchan with rank(EEG.data(:,:)).

Average referencing reduces the rank of the data by 1. When you rereference to average,  the sum of all your channels is 0 at each time point.  One channel is therefore a linear combination of the others, which implies statistical dependence, therefore the dimensionality of your data is reduced  by one after averaging. I always average reference after ICA.

Average referencing before means you can only extract N-1 independent components after this.  This is fine as long as you remember to extract 1 component less than you would have before average reference.

Best wishes,

Jumana


From: eeglablist-bounces at sccn.ucsd.edu<mailto:eeglablist-bounces at sccn.ucsd.edu> [mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Eric HG
Sent: 08 December 2016 00:39
To: eeglablist <eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>>
Subject: [Eeglablist] Artifacts after removal of ICA component

Hi everybody,

I've had some trouble with removal of components using ICA (I'm currently using ICA infomax). After I remove a component there seems to be induced high frequency noise in the dataset.

The data is referenced to average before applying ICA.

Does anyone know why that is happening?

Best,

Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20161210/3c1b9e74/attachment.html>