freqstatistics

[Michael Wibral]

Hi Patricia,

from looking at your plots I see that you average over the whole frequency range analysed (40-180). Only a few frequency bands seem to carry a significant effect - hence the small units, when displaying the mask (0.0426 is the maximum already!). I suggest you try to indetify the frequency band with the geratest effect and then compare stat/mask/statmask again. In the statmaskplot you're amplifying the effect by multiplying mask and (tiny) effects.

What might seem odd for you is that some sensor has a high t-value (averaged over frequencies) while it has a very low value in the average mask. But this can happen:
Imagine your frequencies being:
f = [40 42 ...178 180]      (70 entries)

the t-values at that sensor for each frequency are
stat(sensor,:) = [ 0 0 ....... 0 70]    (70 entries, one for each frequency, only one is nonzero)
The average over all frequencies (which you plot) in this case is 70/70=1

The mask will be
mask(sensor,:) = [0 0 ....... 0 1] (70entries, one for each frequency, only one is nonzero)
he average over all frequencies (which you plot) in this case is 1/70 i.e. a tiny value!

Therefore plotting t-stats averaged over frequencies and the mask averaged over frequencies may give very different results.

Michael




-----Ursprüngliche Nachricht-----
Von: Patricia Wollstadt 
Gesendet: Apr 7, 2010 3:58:20 PM
An: FIELDTRIP at NIC.SURFNET.NL
Betreff: [FIELDTRIP] freqstatistics

>Dear list members,
>
>I am trying to calculate the statistics for four groups in a resting state condition. 
>The problem is that mask and statmask do not match the plotted statistics (see attached figure, topoplotTFR with zparam='stat', zparam='mask', zparam='statmask'). At this point the groups only consist of 8 participants per group and I am now wondering, whether the problem is caused by the low group size or a failure in the skript (I apologize for the quantity of code, but I followed the tutorial very closely and couldn't find an error so far): 
>
>cfg=[];
>cfg.grad=grad; 
>cfg.layout=prepare_layout(cfg);
>cfg.method    = 'montecarlo';
>cfg.channel   = myChannels; 
>cfg.statistic = 'indepsamplesF'; 
>cfg.correctm  = 'fdr'; 
>cfg.numrandomization = 1000;
>cfg.tail = 0;
>cfg.alpha = 0.05;
>cfg.parameter='powspctrm';
>cfg.avgoverfreq = 'no';
>cfg.avgovertime = 'yes';
>
>design = [1:groupSize 1:groupSize 1:groupSize 1:groupSize];             
>design(2,:) = [ones(1,groupSize) 2*ones(1,groupSize) 3*ones(1,groupSize) 4*ones(1,groupSize)];
>
>cfg.design=design;
>cfg.uvar = 1;       
>cfg.ivar = 2;       
>
>cfg.frequency   = [40 180];                 
>statisticsGamma = freqstatistics(cfg,group1avg,group2avg,group3avg,group4avg);
>
>statisticsGamma.statmask=(statisticsGamma.stat).*(statisticsGamma.mask);    
>
>
>Thanks and best regards
>Patricia Wollstadt
>-- 
>GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT!
>Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01
>
>----------------------------------
>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/neuroimaging/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. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Michael Wibral.vcf
Type: text/x-vcard
Size: 628 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20100408/9c698ac2/attachment-0002.vcf>