covariance in timelockanalysis

[Christian Hesse]

Hi Sameer,

> I use the following command to run timelockanalysis to get both
> averages and covariance.
>
> cfgTLA = [];
> cfgTLA.covariance = 'yes' ;
> cfgTLA.covariancewindow = [.100 .300];
> cov_avg = timelockanalysis(cfgTLA,data) ;
>
> (Data actually runs from [-.499 2.0]
>
> This returns
>
> 	      avg: [308x2500 double]
>               var: [308x2500 double]
>           fsample: 1000
>        numsamples: [29x1 double]
>              time: [1x2500 double]
>            dofvec: [1x2500 double]
>             label: {308x1 cell}
>            dimord: 'chan_time'
>               cov: [308x308 double]
>     numcovsamples: [29x1 double]
>              grad: [1x1 struct]
>               cfg: [1x1 struct]
>
>
> however, cov_avg.cov is essentially all zeros. Is there something I
> am missing here?

the scaling in the covariance matrix depends on the scaling of your
signals - if you have the data in units of Tesla (as opposed to
femto) then of course the covariances are going to be very small.
This can cause some covariance based analysis methods to "think" that
your data is rank deficient, etc ...

In any event you can try this and see which message you get:

if (length(cov_avg.label) == rank(cov_avg.cov))
	disp('everything is fine');
else
	disp('you have a problem');
end

If you get the second one you may want to try rescaling your data to
femto Tesla first, that way you get "better behaved" covariance values.

Hope this  helps,
Christian


----------------------------------------------------------------------
Christian Hesse, PhD, MIEEE

F.C. Donders Centre for Cognitive Neuroimaging
P.O. Box 9101
NL-6500 HB Nijmegen
The Netherlands

Tel.: +31 (0)24 36 68293
Fax: +31 (0)24 36 10989

Email: c.hesse at fcdonders.ru.nl
Web: www.fcdonders.ru.nl
----------------------------------------------------------------------




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20070408/2e5517a5/attachment-0001.html>