[Homer-users] Nonparametric analysis

[Jörn M. Horschig]

Dear Cecile.

 

the easiest way for you to go is to download the Matlab toolbox FieldTrip (http://www.ru.nl/donders/fieldtrip/) and convert your data to FieldTrip style, which obeys to this standard:

http://fieldtrip.fcdonders.nl/faq/how_can_i_import_my_own_dataformat

Robert Oostenveld, one of the authors of that paper, is managing the FieldTrip toolbox, and Eric Maris and he wrote the code of the nonparametric cluster-based permutation test for this toolbox. It’s indeed widely used in the MEG/EEG community. The result of the test, however, is a bit tricky to interpret for people naïve to the basics of the test, see:

http://fieldtrip.fcdonders.nl/faq/how_not_to_interpret_results_from_a_cluster-based_permutation_test

 

I would suggest you do all your preprocessing already in Homer, then export your data and convert your final data matrices into FieldTrip style. FieldTrip does not have a user-interface but is more like a script-toolbox, i.e. you would need to how to set up your data, what functions to call, etc. in the command window or in the Matlab editor. FieldTrip has a quite active discussion list (http://fieldtrip.fcdonders.nl/community/) on which you can ask for advise if you get stuck, also just if you have troubles getting your data into the FieldTrip format. you can also unsubscribe again later ;) 

 

As an alternative, as Ted suggested, you can program the nonparametric test yourself, but that would require even more programming skills. I wouldn’t go down the way to make hundreds of copies of your nirs-files, as that would make your file-inventory a lot more confusing. You can rather make in-memory copies in a Matlab script you’re writing. Depending on your Matlab programming skills, this might be the way to go if you really want to understand how these tests work (Ted explained them correctly). The FieldTrip-import as linked above way might be the faster way to go. Anyway, some sort of programming skills is required. 

 

Btw, in MEG and (extracranial) EEG experiments there are not always a lot of trials, but mostly 80-100 per condition are needed for solid effects. That mostly depends on the signal-to-noise ratio, i.e. it depends on the signal you are interested in, and generally the more trials the better the condition-independent noise cancels out. For example, fMRI has a much higher SNR, thus solid analysis can performed on only 20 trials per condition. Depending on what you’re looking for in EEG/MEG studies, 20 trials is usually not sufficient. However, the number of trials does not affect the validity of the nonparametric tests and can be better suited for a low number of trials as they do not require your data to be parametrically distributed (e.g. normal distributed for t-tests). Normally distributed data is much harder to obtain with only a few observation.

 

Best regards,

Jörn

 

--

 

Jörn M. Horschig, Software Engineer

 <http://www.artinis.com/> Artinis Medical Systems  |  +31 481 350 980 

 

From: homer-users-bounces at nmr.mgh.harvard.edu [mailto:homer-users-bounces at nmr.mgh.harvard.edu] On Behalf Of Cécile Issard
Sent: Friday, March 6, 2015 3:14 PM
To: Homer-users at nmr.mgh.harvard.edu
Subject: [Homer-users] Nonparametric analysis

 

Hello Homer users,

 

Is there a way to conduct non-parametrical analysis in homer, similar to those proposed by Maris & Oostenveld (2007) for EEG data ?

 

Best regards,

 

Cécile Issard

Doctorante

01.42.86.43.20

Laboratoire Psychologie de la Perception - UMR8242

45 rue des Sts Pères
75270 Paris cedex 06
 <http://lpp.psycho.univ-paris5.fr/index.php> http://lpp.psycho.univ-paris5.fr/index.php

Labo bébé :  <http://recherche.parisdescartes.fr/LBB> http://recherche.parisdescartes.fr/LBB

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/homer-users/attachments/20150309/d30409e7/attachment.html