[FieldTrip] Power calculations for cluster-corrected stats?

Alik Widge alik.widge at gmail.com
Tue Feb 2 15:59:56 CET 2016


Dear all,

We're working on a plan for a new study in which we plan to be analyzing
EEG in sensor space, taking the Maris/Oostenveld cluster correction
approach since we do not have a clear sense of where our spectral
perturbations will be. Here's the question: has anyone seen a paper talking
about how to do a priori power analysis for this approach? The only thing I
can think to do is guess at an effect size based on similar studies, then
simulate out various sample sizes and trial counts by adding a simulated
signal to some existing EEG dataset.

The problem I'm seeing is that exploring the (nsubject,ntrials) space in
that stimulation approach is going to take quite a while, since it's
multiple instances of an at least 200+ permutation at every possible sample
size combination. More so if I want to explore the consequences of bad
assumptions about effect size.

Anyone know of a more elegant approach? Our literature searches have not
been fruitful.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160202/fd9b5671/attachment.html>


More information about the fieldtrip mailing list