[FieldTrip] distributed computing with montecarlo clusterstatistics

[Nick Ketz]

Hi list, thanks in advance for input that can be provided...

I'm working with a large data set involving all pairwise WPLI estimates 
for a 128 surface  electrode EEG experiment(8128 pairs), and I would 
like to perform a 2 condition, within-subject time-frequency (and space 
if that's possible) cluster analysis.

Given this I have a large number of possible clusters, I have been 
getting on the order of 10,000 observed clusters(with a standard 
clusteralpha of 0.05), which then leads to a lengthy randomization 
test.  What I'm wondering is if there's any way to run this in a 
distributed way(peer-to-peer, or perhaps in parallel)?  I could imagine 
a single processor handling one iteration of the randomization test at a 
time, however when I looked at the statistics_montecarlo.m file I saw no 
quick way to implement this.

I've explored some ideas with 'parfor' loops but I wanted to get the 
lists' input before I start rewriting statistics_monetcarlo.m.  Has 
anyone approached this problem thus far, and if so are there any 
currently available resources?


thanks again...

Nick