[Eeglablist] ICA question

[Arnaud Delorme]

Dear Sarah,

yes, you are correct on both of your points. You may apply the weights calculated on a shorter dataset to a longer dataset and it is (usually) better to apply this procedure than to apply the ICA on the longer dataset with the noise.

Best,

Arno

On 18 Sep 2012, at 10:37, Sara Graziadio wrote:

> Thanks a lot for your answers.
> I wanted just to make sure that I c
> And also to make sure that it is better to apply this procedure than to apply the ICA on the longer dataset with the noise.
> Best wishes
> Sara
>  
> From: Tarik S Bel-Bahar [mailto:tarikbelbahar at gmail.com] 
> Sent: 18 September 2012 04:03
> To: Sara Graziadio
> Cc: eeglablist at sccn.ucsd.edu
> Subject: Re: [Eeglablist] ICA question
>  
> A good strategy is to try to clean your data as much
> as possible of extreme artifacts 
> (these extreme artifacts would 
> attract ICA's attention too much).
>  
> Then, if you have enough time points in your data
> to match ICA's requisites, you should be able
> to achieve a clean and interpretable ICA decomposition.
> With ICs that look interpretable in relation to known
> EEG "components" expected in your protocol.
> You may also find "real" ICs that you did not expect.
>  
> The resultant ICA information can then be 
> transferred to a longer continuous file,
> as you have done. You will however still
> have to deal with artifactual time periods 
> and artifactual epochs, one way or another.
>  
> ICA does not care if you give it epochs or continuous data. However it is important that you feed it 
> enough good clean data, data during which 
> the cognitive behavior you are interested in 
> is occurring. In addition to eliminating 
> "artifactual periods" you may also 
> want to eliminate all periods that are not
> "cognitively"-relevant or "task"-relevant.
>  
> Good luck with your process and let the list know how 
> things go for you.
>  
>  
>  
> 
> On Mon, Sep 17, 2012 at 4:26 AM, Sara Graziadio <sara.graziadio at newcastle.ac.uk> wrote:
> Hello eeglab users,
> I have a question about ICA. My data have some noise in some time intervals. I want to remove the noise before using the ICA but I want to have the whole dataset (continuous data, not epoched) to run some more analysis once the data are cleaned with the ICA. Is there a way to do this?
> At the moment I am running the ICA on the dataset without the noise and then I am applying the ICA weights calculated on the short dataset to the whole dataset (with the noise). Do you think I can do this? Or is there any better method to obtain the whole dataset cleaned without decreasing the ICA performance?
> Thank you
> 
> Best wishes
> 
> Sara
> 
> Sara Graziadio, PhD
> Research Associate
> Institute of Neuroscience
> Newcastle University
> 
> Address:
> Sir James Spence Institute
> Royal Victoria Infirmary
> Queen Victoria road, NE1 4LP
> Newcastle upon Tyne, UK
> 
> Tel:  +44 (0)191 282 1377
> 
> 
> 
> 
> _______________________________________________
> 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
>  
> _______________________________________________
> 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/20120930/c69c872a/attachment.html>