[FieldTrip] stats on planar gradient topographies

"Jörn M. Horschig" jm.horschig at donders.ru.nl
Mon May 16 14:52:24 CEST 2011


Hi all,

First of all, @julian: It would be nice if you wrote in English so that 
all people can understand what you are saying :)

Second, I would like to mention ft_neighbourplot, which is a relative 
new function and plots sensors and its' neighbours in 2D or 3D, 
depending on whether you got 2D or 3D coordinates. Using that function, 
you can verify whether your neighbourselection looks meaningful and 
makes sense.

Finally, we are currently working on changing/improving the 
neighbourselecion function. There will be different methods to choose 
from, and we (me) will create templates for the most common systems that 
can be used instead of relying on a particular mathematical selection 
method. We will announce this when it's done (probably soon).

Maybe Julian is right and Luisa's problem is due to a 'wrong' 
neighbourselection. Given that cfg.minnbhchan is set to 4, this might be 
likely, because not all of the found clusters consist of 4 or more 
channels. What system are you using?

Best,
Jörn


On 5/16/2011 2:32 PM, Julian Keil wrote:
> Hi Luisa,
>
> schnell from the top of my head: wie definierst du denn deine neighbors?
> Ich seh nirgendwo ein Feld cfg.neighbors....
> Ich muss das bei unserem 4D-System immer angeben....
> Es könnte sein, dass du die Struktur, welche Kanäle neben welchen sind 
> falsch definierst, so könntest du zu deinen "random" Clustern kommen.
>
> Ich mache diese Neighbor-Struktur folgendermaßen:
>
> cfg=[];
> cfg.neighbourdist=0.1;
> %cfg.grad=Comb_AV1_1{1}.grad;
> cfg.layout='4D148.lay';
> cfg.channel={'all'};
> %cfg.grad.pnt=cfg.grad.pnt*100;
> dummy=neighbourselection(cfg);
>
> Da bekomme ich auch immer einen Output, wie viele Channels so in der 
> Umgebung sind (sollten in etwa 5 sein).
>
> Hoffe das hilft dir weiter
>
> Viele Grüße
>
> Julian
>
> Am 14.05.2011 um 22:37 schrieb Luisa Frei:
>
>> Hi there,
>> I am trying to compute a simple cluster-based statistic in a between 
>> subjects design. I have followed the steps in the respective tutorial 
>> very closely.
>> This the code I used:
>>
>> %%%%%%%%%%%%%%%%%%%%%%%%%%%
>> cfg = [];
>> cfg.method = 'montecarlo';
>> cfg.statistic = 'indepsamplesT';
>>
>> cfg.correctm = 'cluster';
>> cfg.clusteralpha = 0.001;
>> cfg.clusterstatistic = 'maxsize';
>> cfg.minnbchan = 4;
>>
>> cfg.tail = 0;
>> cfg.clustertail = 0;
>> cfg.alpha = 0.005;
>> cfg.numrandomization = 500;
>>
>> design = [ones(1,13) ones(1,13)+1];
>>
>> cfg.design = design;
>> cfg.ivar  = 1;
>>
>> cfg.channel = {'MEG',sub(1).allbadchan{:}};
>> cfg.latency = [.15 .2];
>>
>> [stat] = ft_timelockstatistics(cfg, grandavgG,grandavgE); % 
>> individuals kept
>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>>
>> My problem is, that I'm not sure which time resolution to use. I 
>> would like to use a small resolution of 10 ms, but this leads to a 
>> fairly random pattern of significant sensors (see attached figure, on 
>> the right). Of course, the more time points I include, the less 
>> significant sensors are left over. However, another problem is, that 
>> the ones that are left (figure, left), are not related to big 
>> differences in the activation pattern at all. How is it possible that 
>> my significant sensors end up being so randomly distributed? Also, 
>> ft_timelockstatistics only comes up with one positive and one 
>> negative cluster., but the sign. sensors are not always connected. 
>> Could anyone give some helpful input here?
>>
>> Thanks,
>> Luisa
>>
>> Fyi: I have tried different thresholds and different numbers of 
>> neighbourhood channels, but there's not much of a change.
>> <Picture 2.png>_______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl <mailto:fieldtrip at donders.ru.nl>
>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip


-- 
Jörn M. Horschig
PhD Student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen

P.O. Box 9101
NL-6500 HB Nijmegen
The Netherlands

Contact:
E-Mail: jm.horschig at donders.ru.nl
Tel:    +31-(0)24-36-68493
Web: http://www.ru.nl/donders

Visiting address:
Trigon, room 2.30
Kapitelweg 29
NL-6525 EN Nijmegen
The Netherlands

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20110516/f1edc2d0/attachment-0002.html>


More information about the fieldtrip mailing list