[FieldTrip] cluster analysis correlation

Ingrid Nieuwenhuis inieuwenhuis at berkeley.edu
Wed Feb 20 00:59:13 CET 2013


Hi all,

I've computer correlations (kendall's tau) between EEG power and a 
behavioral measure over participants (made my own statfun using corr). 
Now I'd like to use the cluster randomization framework to correct for 
multiple comparisons. I think it would be valid to shuffle the 
assignment of EEG power to behavioral measure (which is specified in 
cfg.design for correlation), make a distribution and test my 
correlations against that. So shuffle the data so that say EEG power of 
subect1 get's linked to behavioral measure of subject3, subject2 to 
subject6 etc, etc, just as we do with t-values shuffling who belongs to 
which condition.

Questions:
1) is this implemented in FieldTrip? How would I do this? Cause the 
design needs to be shuffled now, this is different from how it would 
work for a t-test no?
2) If 1== false, Anyone done this, got some tips?
3) Is the above approach I suggest statistically valid? I think testing 
the tau against zero is not valid (seeing if it is significantly 
different from zero by making some dummy stat with zeros), but what I 
suggest is, anyone any thoughts?

Thanks!!
Ingrid

-- 
Ingrid Nieuwenhuis PhD
Postdoctoral Fellow
Sleep and Neuroimaging Laboratory
Department of Psychology
University of California, Berkeley
California 94720-1650
Tolman Hall, room 5305




More information about the fieldtrip mailing list