dics method for coherence analysis

Jan Mathijs Schoffelen Jan.Schoffelen at FCDONDERS.RU.NL
Sat Apr 28 18:32:12 CEST 2007


Dear Feng,



I see that you use the following configuration for your frequency-analysis:
channels with the dics method, following is my script:



cfg = [];

cfg.method='mtmfft';

cfg.ouput='powandcsd';

cfg.keeptrials='yes';

cfg.taper='dpss';

cfg.tapsmofrq=4;

cfg.foilim =[8 12];

cfg.sgncmb = channelcombination({'MEG','MEG'},bldata.label);

blfreq = freqanalysis(cfg,bldata);

clear bldata





This in itself is OK, but given the fact that you use cfg.sgncmb instead of
cfg.channelcmb, you seem to use an old version of the freqanalysis function
(with old syntax, but newer versions are backward-compatible with respect to
these formulations). I'm not sure, but the tutorial documentation on the
fieldtrip-website, as well as the help content of the respective m-file
should contain the most up-to-date info about the proper configuration
terminology (provided you have the most up-to-date version of fieldtrip). I
would advise anyhow to get the latest version (see below as well).



For the sourceanalysis you use the following configuration:



cfg = [];

cfg.grid = source2sparse(ldf);  % Only consider gridpoints within head
volume

cfg.method = 'power';

cfg.refdip=refdip;

cfg.numrandomization = 100;

cfg.randomization = 'yes';

cfg.projectnoise = 'yes';

cfg.lambda = 0;

cfg.hdmfile = headmodel;

cfg.frequency = (foi(1)+foi(2))/2;

cfg.keepleadfield = 'no';

cfg.feedback = 'none';

[sourceRand] = sourceanalysis(cfg, eoifreq, blfreq);



I don't see why this shouldn't work but it is computationally very demanding
(as you might have noticed), because you recompute for each randomization
two volumes based on two sets of newly computed spatial filters (for each of
the shuffled conditions). Moreover, in your comparison between the baseline
and your epoch of interest (of your observed data), you compute two sets of
spatial filters as well. Our experience however is, that source
reconstruction is more robust (and computationally less expensive, and
interpretation-wise less complicated) when you use common filters, i.e.
collapsing the cross-spectral densities for the baseline and the active
condition (even though there might be a clear difference between the
conditions, e.g. dipoles switching on or off).

Either way, you betray yourself using an old version of fieldtrip by
specifying cfg.method = 'power', which should be cfg.method = 'dics'.



cfg=[];

cfg.comparestat = 'relchange' ;

cfg.method = 'randomization' ;

cfg.parameter = 'coh'  ;

cfg.approach = 'parametric';

sourcestat = sourcestatistics(cfg,sourceRand);



Again, this configuration is a relic from old times, and you should use the
function sourcedescriptives to get what you want (given your preceding
steps). I guess, that your variable sourceRand contains the fields avgA,
avgB, and trialA, and trialB. If you look into the code (of the latest
sourcedescriptives.m) you should be able to get an idea of what's going on
there.



I found in the output variable sourceRand subfield 'pow' and 'coh', which
one shall I use for statistical analysis? I assume it should be 'coh', then
what does 'pow' stand for? Is it the psd with the target frequency range for
each virtual channel? Then will the 'pow' values be different from the
computation without specifying reference virtual channel in the script?



Indeed, pow stands for the power at the specified voxels. This shouldn't
change when you use a reference channel.



Yours,



Jan-Mathijs

_______________________________________________ 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.
http://listserv.surfnet.nl/archives/fieldtrip.html
http://www.ru.nl/fcdonders/fieldtrip/

----------------------------------
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.
  http://listserv.surfnet.nl/archives/fieldtrip.html
  http://www.ru.nl/fcdonders/fieldtrip/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20070428/c32a6879/attachment-0001.html>


More information about the fieldtrip mailing list