Freqanalysis - clusterthreshold and neighbourselection

Eric Maris e.maris at DONDERS.RU.NL
Fri Oct 2 11:51:06 CEST 2009


Hi Nina,









I have two questions regarding clusteranalysis.



1.	When running 'freqstatistics' on time frequency/frequency data with
the method 'montecarlo' which clusterthreshold would you use? I would think
that with the data usually not being normally distributed you would use a
nonparametric threshold. However, the default value for clusterthreshold is
'parametric'.

a.	Is there documentation about what 'nonparametric_individual' and
'nonparametric_common' mean? If yes, where can I find it?



You can safely use the parametric threshold, regardless of the probability
distribution of your data. The reason is that this threshold only affects
the type of effect for which you test statistic is sensitive. The false
alarm rate does not depend on it. See Maris & Oostenveld (2007).



You only use the two other clusterthreshold options if you use a
non-normalized channel-level statistic. If you construct clusters by
thresholding channel-level T-statistics, go ahead with the parametric
threshold.





2.	I am looking at data from our old Neuromag 122 system and used the
function 'neighbourselection' to define neighbouring channels. When checking
what 'neighbourselection' did I found that the neighbours identified were
not only the sensors surrounding the sensor of interest, but also sensors
lying far away from it. Also, not all sensors surrounding the one of
interest were included as neighbours. I used the gradfile generated by
'mne2grad122' as cfg.grad and different values for neighbourdist (4 and 3).
Does anybody have a solution for this?





I guess Neuromag 122 system has planar gradiometers. For this sensor
configuration, the definition of a neighbourhood-metric requires some
thinking (because you are actually measuring spatial gradients with
different orientations). I contributed to a recent thread on the Fieldtrip
Discussion list about cluster-based permutation tests for the Vectorview
system. If I'm not mistaken, a scientist from an imaging center in Paris
initiated this thread. Have a look in the archive of the Fieldtrip
Discussion list (keywords: planar, cluster, Vectorview).





Good luck,











Thanks in advance for your help!



Nina



----------------------------------

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.

http://listserv.surfnet.nl/archives/fieldtrip.html

http://www.ru.nl/fcdonders/fieldtrip/


----------------------------------
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/20091002/15335b00/attachment.html>


More information about the fieldtrip mailing list