I have promising but inconclusive FEAT results from scanning two subjects.
I'm using cluster thresholding since the effects are too small for a
voxel-based approach. I would like to have some idea of how many subjects I
need to have a reasonable chance of seeing a significant effect. Does
anyone have suggestions of how to approach this power analysis, specifically
for cluster thresholding? I can't seem to find documentation on exactly how
the clusters are generated or (more importantly) how likely they are under
the null hypothesis.
Many thanks,
Steve
Steven Lowen
Brain Imaging Center
McLean Hospital
115 Mill St
Belmont, MA 02478
http://cordelia.mclean.org/~lowen/
|