[Eeglablist] How do I define rank number for runica() ?
mmiyakoshi at ucsd.edu
Tue Dec 1 12:32:37 PST 2015
Dear Imtiaz and Tatu,
Sorry for the trouble.
If you provide whole length of data, rank() returns very low numbers.
However, if you chop up the data into short epochs such as 1 or 2 seconds
and rank() one by one, it usually returns the number of channels (if your
channels did not bridge).
Looks like runica() runs rank() function as rank(data) which may return
very low rank under certain situation (at least from my experience).
I heard someone say that after low-pass filtering the data rank could be
estimated lower. This seems reasonable speculation, but I've never tested
it myself. I don't know the mechanism of the problem either.
Anyway, it's wroth trying to file it to EEGLAB bugzilla.
Thank you for cooperation.
On Tue, Dec 1, 2015 at 1:04 AM, Tatu Huovilainen <
tatu.huovilainen at helsinki.fi> wrote:
> Rank is an estimate of linearly independent rows or columns of the matrix.
> This estimate is necessary as ICA can't return more components than the
> rank of the input dataset. You can just use
> rank(EEG.data([eeg_channel_indices],:)) in matlab, also check 'help rank'
> or for exact way the rank is calculated 'edit rank'. Trying to pull more
> components than the rank results in components with complex numbers or the
> algorithm not converging.
> http://sccn.ucsd.edu/wiki/Chapter_09:_Decomposing_Data_Using_ICA and the
> next chapter is a good overview of the method.
> Your lower rank can result, atleast, from including the reference channel,
> interpolated channels or bridged channels during measurement.
> Hope this helps,
> On Mon, 30 Nov 2015 10:30:52 -0600
> Imtiaz SA <imtiaz1877 at gmail.com> wrote:
> > When I execute ICA it prompts for rank number and suggests a rank number
> > automatically. I have total 38 channels and for some data sets EEGLAB
> > suggests rank number 21 or 29 or 33 (always less than 38). Could you
> > tell me how exactly these rank numbers are calculated?
> > --
> > Thanks and regards,
> > *Imtiaz*
> > PhD Student
> > Department of Computer Science
> > The University of Texas at San Antonio
> Tatu Huovilainen <tatu.huovilainen at helsinki.fi>
> 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
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the eeglablist