Dear Christian and Dr. Stefan Golaszewski,
> How big the group have to be to need a bonferoni-correction? Do we need
> the bonferoni correction for a group with 10 samples?
The Bonferroni-correction is related to the number of tests performed (ie.
the number of voxels or resels) rather than the number of samples (ie.
scans). Whether you have 10 or 100 samples would not matter.
In SPM99 both the uncorrected and the (resels) Bonferroni-corrected
p-values are reported as well as set and cluster p-values, so you would
not have to perform the Bonferroni-correction yourself, - it is already
done.
You might want to Bonferroni-correct the reported p-values from SPM99 if
you test several hypotheses (as written in spm_conman.m).
Finn
__________________________________________________________________
Finn Årup Nielsen, IMM DTU, Lyngby, Denmark
http://hendrix.imm.dtu.dk/staff/fnielsen/
__________________________________________________________________
|