[FieldTrip] Computing cluster sizes on group-level topographic maps without using built in monte carlo statistics

Andrew Heusser andrew.heusser at gmail.com
Thu Jul 17 20:15:54 CEST 2014


Dear Fieldtrippers,
I am working on an MEG analysis where I compute average oscillatory power
for a given band for each trial in my experiment, and then perform a
parametric regression over trials to obtain a t-statistic representing the
fit to the model at each sensor and for each subject (for a given band).
 This leaves me with a topographic map of t-statistics for each subject for
a given frequency band.  Then, to compute reliability across subjects, I
perform a one-sample t-test on the model fits across subjects for a given
sensor to get a group-level topographic map of significance values.

I would like to cluster correct these group-level maps by iteratively
shuffling trials within subject and recomputing model fits, recomputing the
group maps, and then finding the size of clusters to build a null
distribution of cluster sizes.

1) Using the Fieldtrip functions (i.e. ft_freqstatistics), is there a
simple way to grab cluster sizes from these 'shuffled' group-level
statistical maps so that I can build a null distribution of cluster sizes
and find a cluster threshold? Rather, is it possible to obtain cluster
sizes on any statistical map without using the monte carlo statistics?

2) Does this approach logically make sense, or is there maybe another way
to achieve this that I haven't thought of?

Thank you in advance for you help!

-- 
Andy
Graduate Student at NYU
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20140717/d2cfe438/attachment-0001.html>


More information about the fieldtrip mailing list