[Eeglablist] Can I apply cluster weights to all EEG data?

[Makoto Miyakoshi]

Dear Rachel,

> My problem is that the clusters which are most interesting (i.e., they
look like relevant ERPs and contain the most ICs) contain lots of ICs from
some participants and none from others. I assume this must be the case in
any study with a between subjects factor.

This can happen in all within-subject conditions as well. To address the
problem of too many/too few subject per cluster, Nima Begdely-Shamlo
developed measure projection.
http://www.sciencedirect.com/science/article/pii/S1053811913000876

> A suggestion from my PhD supervisor was to find the weights for the
cluster and then apply these weights to the EEG data for every participant
and every condition. That way I can compare these new component activations
across my conditions/subjects knowing that the same IC component is being
compared.

I did not understand what you do.
Would you detail it?

As a general solution I would recommend that you use small number of
clusters (10-15) if you want to include as many unique subjects as possible.

Makoto

On Thu, Jun 26, 2014 at 4:24 AM, Cooper, Rachel <rcoopea at essex.ac.uk> wrote:

> Hi all,
>
> This is partly a theoretical question and partly a practical one. I have a
> study with 2 participant groups and 3 repeated measures factors. So far I
> have run ICA on my epoched data and then used cluster analysis. My problem
> is that the clusters which are most interesting (i.e., they look like
> relevant ERPs and contain the most ICs) contain lots of ICs from some
> participants and none from others. I assume this must be the case in any
> study with a between subjects factor.
>
> A suggestion from my PhD supervisor was to find the weights for the
> cluster and then apply these weights to the EEG data for every participant
> and every condition. That way I can compare these new component activations
> across my conditions/subjects knowing that the same IC component is being
> compared. Is it safe to do this theoretically?
>
> On a practical level, I have found the data used to plot the cluster scalp
> map in STUDY.cluster(clust).topo. However, I don't know how I would use
> this to apply cluster weights to the rest of my data as it is not a matrix
> of channels x ICs like the regular ICA weights. Is there another way to
> access the weights for a cluster?
>
> Any help is much appreciated and if you need more detail please let me know
> This is a fantastic list!
>
> Rachel
>
> Rachel Cooper
> PhD researcher
> Department of Psychology,
> University of Essex,
> Wivenhoe Park,
> Colchester,
> Essex,
> CO4 3SQ
> _______________________________________________
> 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
>



-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20140627/8f89b95e/attachment.html>