[Eeglablist] Remove or retain low-mutual information components in ADJUST?

[Marco Buiatti]

Dear Brian,

while ICs low in the list might represent just noise, they might also
be associated with low-variance artifacts like transitory jumps in one
specific electrode, and these are the ones that ADJUST typically
captures at the bottom of the list.

In any case, since they represent low portions of the overall
variance, I do not think that if they are just noise, eliminating them
would be a problem, at least for a standard ERP analysis.

Do however check that ADJUST works properly on your data before
blindly apply it.

Best,

Marco (ADJUST author).

On 15 January 2014 23:43, Erickson <ericksonb.eng at gmail.com> wrote:
> List,
> I am creating a data processing pipeline using ADJUST to automatically
> reject components. Before I let this monster loose on my data, I want to
> make sure that I should trust ADJUST when it tells me that components low on
> the list are artifacts. To explain:
>
> ICA components are returned in order of decreasing mutual information. Thus,
> component 1 has the highest MI, 2 the next highest, etc. At the EEGLAB
> conference this year, Scott made the point that components very low on the
> list contain very low MI and may not actually represent sources, but noise
> not otherwise accounted for. Yet ADJUST does not seem to have any cutoff -
> it will happily label component 84 out of a dimensionality of 84 as an
> artifact. If this component is not reliable, I certainly don't want to
> eliminate it from my data as an artifact.
>
> Has anybody dealt with this issue? Perhaps by setting some cutoff (half the
> dimensionality? or some measure of MI?) below which we no longer accept that
> calling a component artifactual makes sense?
>
> Thank you,
> Brian
>
> Brian Erickson
> Ph.D. Candidate
> Applied Brain and Cognitive Sciences Program
> Drexel University, College of Arts and Sciences, Department of Psychology
>
> _______________________________________________
> 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



-- 
Marco Buiatti, PhD

CEA/DSV/I2BM / NeuroSpin
INSERM U992 - Cognitive Neuroimaging Unit
Bât 145 - Point Courrier 156
Gif sur Yvette F-91191  FRANCE
Ph:  +33(0)169.08.65.21
Fax: +33(0)169.08.79.73
E-mail: marco.buiatti at gmail.com
http://www.unicog.org/pm/pmwiki.php/Main/MarcoBuiatti

***********************************************