Fwd: [FIELDTRIP] source analysis example images

[Robert Oostenveld]

Hi Brian

On 7 May 2007, at 15:04, Brian Roach wrote:
>> I am wondering if this type of result seems reasonable (if it
>> looks as though I have not made any huge mistakes)?  If so, I
>> would run another condition for the same subject and try to use
>> the statistics functions.
>>
>> Also, what does this color scale describe?  What are its units?

The ratio between projected power and projected noise, where the
noise estimate is based on the smallest singular value of the cross-
spectral density matrix. See van Veen 1997. It is dimensionless. Note
that the magnitude strongly depends on the noise estimate.

But given that you are able to estimate NAI with EEG without
specifying cfg.lambda makes me suspicious: The following code from
"private/beamformer_dics.m" (around line 164) should apply

if projectnoise
   % estimate the noise power, which is further assumed to be equal
and uncorrelated over channels
   if isrankdeficient
     % estimated noise floor is equal to or higher than lambda
     noise = lambda;
   else
     % estimate the noise level in the covariance matrix by the
smallest singular value
     noise = svd(Cf);
     noise = noise(end);
     % estimated noise floor is equal to or higher than lambda
     noise = max(noise, lambda);
   end
end

You did not specify lambda, hence the mallest singular value of the
cross-spectral density matrix will be used (second part of the code).
But in case of averaged referenced EEG, that should be zero. Hence
the noise estimate is zero, resulting in NAI = source power divided
by zero, which should be infinite.

You should ensure that your EEG data is average referenced over teh
channels on which you want to base your source reconstruction prior
to computing freqanalysis and sourceanalysis. If the EEG data is
average reference and you specify cfg.lambda=0 in sourceanalysis, you
should see a warning about the cross-spectral density matrix being
rank deficient and your nai should be infinite. Hence you have to
specify a manual estimate of the noise floor (in cfg.lambda).

Robert

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/fcdonders/fieldtrip.