Freqstatistics actvsblT vs. depsamplesT

[Ingmar Schneider]

Dear fieldtrip community,

I am currently working on an activation vs. baseline comparison script  
(attached below) using the 'freqstatistics' function provided by  
fieldtrip. According to the instructions given in the fieldtrip  
tutorial, I used the 'actvsblT'-statfun for the comparison of 2s long  
data segments over a frequency range from 6Hz to 100Hz. However, this  
turned out to be almost impossible due to huge RAM requirements and  
frequently led to an 'Out of memory-error' even on a machine with 32GB  
of RAM.
In order to make it work, I had to narrow the frequency range and  
number of randomizations. When I used the 'depsamplesT'-statfun for  
the comparison instead of 'actvsblT', a comparison across the whole  
frequency-range was possible requiring much less RAM (~2GB). The  
resulting statistics, however, slightly varied from those obtained  
using the 'actvsblT'-statfun (Topoplots attached); the localisation  
seems to be the same, but the t-values are a little lower.

As far as I could find out using the documentation both statfuns  
calculate dependent samples T-statistics. Is there an essential  
difference between them causing the differences in the statistics?

With best regards,
Ingmar


% Preparation of the statistical design
nsubjects   = size(TFact, 2);
statdesign  = zeros(2,2*nsubjects);
statdesign(1,1:nsubjects)               = 1;
statdesign(1,nsubjects+1:2*nsubjects)   = 2;
statdesign(2,1:nsubjects)               = [1:nsubjects];
statdesign(2,nsubjects+1:2*nsubjects)   = [1:nsubjects];

% Realigning time
TFblavg.time   = TFactavg.time;

% Freqstatistics
cfg = [];
cfg.frequency   = [9 11];
cfg.channel     = {'MEG', '-MLP12', '-MRC14', '-MLT41', '-MRC25',  
'-MRP56', '-MRT21', '-MLO21'};
cfg.latency     = [0.5 2.5];
cfg.parameter   = 'powspctrm';
cfg.method      = 'montecarlo';
cfg.statistic   = 'depsamplesT' / 'actvsblT' % respectively
cfg.correctm    = 'fdr';
cfg.alpha       = 0.05;
cfg.numrandomization    = 1500;
cfg.grad        = grad;
cfg.design      = statdesign;
cfg.uvar        = 2;
cfg.ivar        = 1;
freqstatACTvsBL = freqstatistics(cfg, TFactavg, TFblavg);


-- 
Ingmar Schneider

Max-Planck-Institut für Hirnforschung
Deutschordenstraße 46
D-60528 Frankfurt/Main

Tel.: 069/6301-83221
Fax: 069/96769-327
Mail1: schneider at mpih-frankfurt.mpg.de
Mail2: ingmar.schneider at bio.uni-giessen.de

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

----------------------------------
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: FS_actvsblT_9-11Hz.jpg
Type: image/jpeg
Size: 145789 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20090623/6cf0632b/attachment-0002.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: FS_depsamplesT_9-11Hz.jpg
Type: image/jpeg
Size: 144553 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20090623/6cf0632b/attachment-0003.jpg>