[Eeglablist] How many clusters to choose for a STUDY?

[Aleksandra Vuckovic]

Dear Joaquin,
that has lots of sense, but if you 130 IC in one cluster, is that still OK or an indicator that larger number of clusters is needed? Would you recommend increasing number of clusters all until each cluster contains at least 1 IC from each subject? My experience is that this results in generating similar clusters.
Also I have noticed that cluster dipole gives deeper location of sources then I would expect. Is that right?
Regards,
Alex
.________________________________________
From: Joaquin Rapela [rapela at ucsd.edu]
Sent: 05 March 2012 23:29
To: Aleksandra Vuckovic
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] How many clusters to choose for a STUDY?

Dear Aleksandra,

I select the number of clusters in such a way that (ideally) each cluster has one component from every cluster. That is if my study contains 27 subjects, I select the number of subjects so that each cluster contains 27 components from 27 subjects. Of course, this is only an ideal scenario, but one that could guide you to a good number of clusters.

After you have decided on a number of clusters, and analyzed your data with these number of clusters, it is a good practice to repeat the analysis with a slightly different number of clusters, to get an idea of the robustness of your conclusions.

Cordially, Joaquin

On Mon, Mar 05, 2012 at 09:41:08PM +0000, Aleksandra Vuckovic wrote:
> Dear all,
> I’m clustering ICAs of three groups in a STUDY and was just wondering what would be the best indicator for how many clusters are just right (apart for experimentally testing different numbers). I’ve noticed that some cluster contain more than 100 components while some other 20-30 IC . Does it mean that this with 100 component is too large so I should go for larger number of clusters to separate this cluster in two, or 20-30 CI is too small number and I should reduce the total number of clusters?
> Many thanks,
> Aleksandra

> _______________________________________________
> 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

--
Joaquin Rapela, PhD
Swartz Center for Computational Neuroscience
University of California San Diego
9500 Gilman Drive,
San Diego, CA 92093-0559
tel: (858) 822-7536
fax: (858) 822-7556
http://sccn.ucsd.edu/~rapela
----------------------------------

Yet who reads to bring an end however desirable? Are there not some pursuits
that we practice because they are good in themselves, and some pleasures that
are final? And is not this among them?  I have sometimes dreamt, at least, that
when the Day of Judgment dawns and the great conquerors and lawyers and
statesman come to receive their rewards--their crowns, their laurels, their
names carved indelibly upon imperishable marble--the Almighty will turn to
Peter and will say, not without a certain envy when He sees us coming with our
books under our arms, ``Look, these need no reward. We have nothing to give
them here. They have loved reading.''
                                                           Orlando. A biography
                                                           Virgina Woolf