Cluster change over time

[Eric Maris]

Dear Rodolphe and Michael,

 

 

strictly speaking cluster analysis (as a construct to get a univariate test instead of a mass-univariate one) only tells you that there IS a significant cluster. The exact shape and temporal evolution of this cluster is not part of what's tested - as least as far as I understand this issue. That is a data point can surpass your cluster inclusion threshold by chance - a portion of alpha data points will actually do this. In addition, these point can be accidentally close to other clustermembers. Hence, it will be included in a cluster. So 'true' cluster membership of a certain electrode/timepoint/frequency is not statistically 'guaranteed'. Additional tests would be necessary.
Then one thing that this argument does not account for is that 'accidentally neighbouring a cluster' is not so likely because of the smoothness of MEEG data - but I do not know at the moment how you could exploit this.




This is well formulated. The null hypothesis that is tested by the cluster-based permutation tests involves that the spatiotemporal biological data in the experimental conditions that are being compared are drawn from the same probability distribution. They are NOT testing null hypotheses of the following type: there is no difference between the conditions at channel A and time point T whereas there may be differences at other (channel, time point)  pairs. In fact, such null hypotheses do not make much sense when dealing with electrophysiological data because (a) because of volume conduction, spatially localized effects are not plausible, and (2) because neurophysiological processes are not instantaneous, highly temporally localized effect are not plausible either.

 

Best,

 

Eric Maris

 

 

 

 

 

 

 

Michael

  _____  

Von: Rodolphe <batrod at GMAIL.COM>
Gesendet: Oct 20, 2010 12:49:06 AM
An: FIELDTRIP at NIC.SURFNET.NL
Betreff: [FIELDTRIP] Cluster change over time

Dear Fieldtrip users,  

I used the Monte-Carlo method with cluster correction to compare two conditions within the same group of subjects.

I didnt average over time, thus the clusterplot showed several topographic plot of the cluster that was found to be significant.

I can see that this cluster evolves over time (number and place of significant electrodes highlighted). My question is, considering the statistical method using for this analysis, can i say that this change over time is significant or does it require further analysis like ANOVAS in concerned electrodes?

Thanks a lot, 

Rodolphe N., Ph.D.

--------------------------------------------------------------------------- You are receiving this message because you are subscribed to 
the FieldTrip list. 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.
---------------------------------------------------------------------------

--------------------------------------------------------------------------- You are receiving this message because you are subscribed to 
the FieldTrip list. 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.
---------------------------------------------------------------------------


---------------------------------------------------------------------------
You are receiving this message because you are subscribed to
the  FieldTrip list. 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 --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20101021/db262138/attachment-0002.html>