[FieldTrip] Neuromag sensor level stats

Peter Goodin pgoodin at swin.edu.au
Tue Sep 25 02:37:14 CEST 2012

Thanks very much for the responses Stephan and Jörn!

One last question if possible Jörn. In your opinion, will doing several (at the most 5) cluster based t-tests affect the validity of permutation method testing to control for multiple comparisons? Would being more conservative regarding my alpha (for example a combination of permutation AND Bonferroni) possibly alleviate this?

Thanks again for the responses,


From: fieldtrip-bounces at science.ru.nl [mailto:fieldtrip-bounces at science.ru.nl] On Behalf Of "Jörn M. Horschig"
Sent: Monday, 24 September 2012 10:15 PM
To: FieldTrip discussion list
Subject: Re: [FieldTrip] Neuromag sensor level stats

Dear Peter,

regarding your first question, most people here at the Donders actually repair bad channels. When computing the grandaverage over subjects or doing some statistics across subjects, only those channels which are in common for all datasets are used. Thus, if having different channels damaged/rejected per subjects, you can end up with a fairly low amount of channels. But, in the end it depends pretty much on what channels you reject wrt your hypothesis, e.g. having few occipital channels while being interested in motor regions won't constitute a big problem, so no need to interpolate. I leave the question whether it's a correct thing to do up to you (imho: I wouldn't worry too much about it).

Regarding your third question: In FieldTrip you can only compare two conditions directly (at least when doing cluster-based permutation testing). Also, any interaction needs to be setup manually.


On 9/23/2012 3:14 PM, Peter Goodin wrote:
Hi Fieldtrippers,

I'm at a stage where I'd like to do statistical analysis on my ERF neuromag data, but don't really know where to begin. I'd like to do a cluster analysis, but the tutorial material is for a CTF system. I'm using a neuromag 306 channel which complicates matters somewhat due to the different sensor types. I've searched the mailing list but can't find any solid answers / example scripts.

The most pressing questions I have are:

1. Despite running the data through maxfilter and entering in the bad channels, occasionally one gets through (not the same one). I've used Fieldtrip's repairchannel function on the data, making sure to replace the channel only with like sensor neighbouring sensors. While I know this is a common method for fixing bad channels in EEG, does the same hold for MEG due to non-existent smearing of the signal? Is it best to keep the fixed channel(s) or just remove them from the stats analysis across all participants?

2. Given the three sensor types on the neuromag system, do I need to run three different cluster analyses (one for mags, one for xgrads, one for ygrads), each time specifying like sensor neighbours? Alternatively, If I'm not interested in the mags data, can I just  run one using the data from the combined gradiometers?

3. When examining for group x condition effects, would I use a three row design matrix with row 1 containing group, row 2 containing participants and row 3 conditions?

Thanks for any help,



fieldtrip mailing list

fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>



Jörn M. Horschig

PhD Student

Donders Institute for Brain, Cognition and Behaviour

Centre for Cognitive Neuroimaging

Radboud University Nijmegen

Neuronal Oscillations Group

FieldTrip Development Team

P.O. Box 9101

NL-6500 HB Nijmegen

The Netherlands


E-Mail: jm.horschig at donders.ru.nl<mailto:jm.horschig at donders.ru.nl>

Tel:    +31-(0)24-36-68493

Web: http://www.ru.nl/donders

Visiting address:

Trigon, room 2.30

Kapittelweg 29

NL-6525 EN Nijmegen

The Netherlands
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20120925/8f9b2726/attachment-0002.html>

More information about the fieldtrip mailing list