I found Matthew Brett et al's description of Random Field Theory, FWE, Euler characteristics, etc. helpful:
I would also like to second Marko's suggestion to look into nonstationary smoothness correction (for cluster-level probabilities). Check out archive posts on this topic and papers by Hayasaka and Worsley-- also Moorhead Neuroimage 2005. My group uses the stat_threshold script from Worsley's fmristat package to do the nonstationary smoothness correction.
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of J Li
Sent: Tuesday, January 13, 2009 9:01 AM
To: [log in to unmask]
Subject: [SPM] VBM2 - Correction level problem
I am a postgraduate student doing my thesis on the comparison of the gray
matter volume among boys and girls.
I am using the Christian Gaser VBM2 toolbox on SPM2 running on Matlab
6.5.1..I have tried to use the mni template and also the customized children
template with all the default procedures in the VBM toolbox.
I use the age and GM volume as the covariates in the Anova model provided in
the toolbox. When I set the result as uncorrected and p=0.001, it shows some
more-likely results comparing two groups, however, when I set it into FDR or
FWE, there are no results.
As I know after doing FWE/FDR correction, the results should be more rigid, so
it may lead to less significance of my data.
Am I know do I put any wrong steps so far? Do anyone know that what the
FWE/ FDR correction do inside the calculation? What factors are able to affect
the results after doing the FDR/FWE ?