Update: Freqstatistics Now Yields (Some) Significant Clusters

Charles Cook charles.cook at ULETH.CA
Wed Jun 24 16:22:04 CEST 2009


Hi Michael,

I would say the single most important variable was getting the scale
correct, something you suggested in your last post. Thus, getting this
correct in our code:

%Read in the electrode locations for the Std81 montage
cfg = [];
elec = read_fcdc_elec('EGI-BESA_Standard_81_prime.sfp');
elec.pnt = 1000*elec.pnt;

This really set us on the right track. Also, having the correct neighbour
distance was quite important. That took a little bit of tweaking to get right. 

Having said all of this, the cluster analysis is proceeding on our data. One
point we're not completely satisfied with is the within subjects analysis
using the dependent t-test. In our study, we are interested in how male
participants were performing in task1 vs. task2. Behaviourally we do see a
significant RT difference between task1 vs. task2, but when we run the
analysis, we are unable to find any significant clusters (<0.05). It may be
possible that this method simply might be too conservative, so if anyone has
any suggestions, I'd be very interested and appreciative in some feedback.

Cheers,

Charles Cook
==============================================

% perform the statistical test using randomization and a clustering approach
% using the NEW freqstatistics function
cfg = [];
cfg.neighbourdist    = 45;
cfg.elec             = elec;
cfg.statistic        = 'depsamplesT';
cfg.minnbchan        = 0;
cfg.clusteralpha     = 0.05;
cfg.alpha            = 0.05;
cfg.clustertail      = 0;
cfg.numrandomization = 10000;

cfg.latency          = [0 1000];
cfg.frequency        = [8 10];
cfg.avgovertime      = 'no'; 
cfg.avgoverfreq      = 'yes';
cfg.avgoverchan      = 'no';

cfg.correctm         = 'cluster';
cfg.method           = 'montecarlo';
cfg.feedback         = 'gui';
%cfg.design           = [1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
21 22;   % subject number
                        %1 1 1 1 1 1 1 1 1 1  1  2  2  2  2  2  2  2  2  2 
2  2]; % condition number

cfg.design           = [1 2 3 4 5 6 7 8 9 10 11 1  2  3  4  5  6  7  8  9 
10 11;   % subject number
                        1 1 1 1 1 1 1 1 1 1  1  2  2  2  2  2  2  2  2  2  2
 2]; % condition number
                          
                    
cfg.uvar = 1;                                   % "subject" is unit of
observation
cfg.ivar = 2;                                   % "condition" is the
independent variable
%stat = freqstatistics(cfg, maleloc_all, femloc_all);
%stat = freqstatistics(cfg, femfeat_all, malefeat_all);
stat = freqstatistics(cfg, maleloc_all, malefeat_all);
%stat = freqstatistics(cfg, femfeat_all, femloc_all);

cfg = [];
cfg.elec  = elec;
cfg.rotate = 0;
%cfg.zlim = [-6 6];
cfg.alpha = 0.05;
cfg.label = stat.label;
cfg.electrodes = 'labels';
cfg.showxlim = 'yes';
cfg.showzlim = 'yes';
cfg.showylim = 'yes';
clusterplot (cfg, stat);
===========================
Obtaining the electrode configuration from the configuration.
there are on average 4.7 neighbours per channel
using "statistics_montecarlo" for the statistical testing
using "statfun_depsamplesT" for the single-sample statistics
constructing randomized design
total number of measurements     = 22
total number of variables        = 2
number of independent variables  = 1
number of unit variables         = 1
number of within-cell variables = 0
number of control variables      = 0
using a permutation resampling approach
repeated measurement in variable 1 over 11 levels
number of repeated measurements in each level is 2 2 2 2 2 2 2 2 2 2 2 
computing a parmetric threshold for clustering
estimated time per randomization is 0 seconds
found 5 positive clusters in observed data
found 12 negative clusters in observed data
using a cluster-based method for multiple comparison correction
the returned probabilities and the thresholded mask are corrected for
multiple comparisons
There are 0 clusters smaller than alpha (0.05)

On Wed, 17 Jun 2009 09:24:41 +0200, Michael Wibral
<wibral at BIC.UNI-FRANKFURT.DE> wrote:

>Hi Charles,
>
>the plots look OK to me. Could you let us know what finally made your
analysis work (if there was a single most important thing)
>- that would be most helpful. 
>
>Thanks,
>Michael
>

----------------------------------
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.



More information about the fieldtrip mailing list