If you have run an unmodified version of the script, the "corrp"
output IS already cluster thresholded (using TFCE) corrected for
multiple comparisons. That's why it is called "corrp" = corrected p.
TFCE has the advantage that no a priory cluster forming threshold
needs to be specified, and it is quite sensitive. If you took the
notion of correcting for multiple testing really seriously, you
probably needed to divide your p-value by the number of components and
the number of directional tests for each component...Therefore, there
is no need (and probably no use) in cluster correcting the uncorrected
data. You could apply a FDR correction, however. See the online manual
on how to do it. On a side note: it is not expected for every map to
behave differentially across groups.
Cheers,
Cornelius
On Wed, Jul 6, 2011 at 6:30 PM, Rosalia Dacosta Aguayo
<[log in to unmask]> wrote:
> Hi,
> I have done dual_regression and I have seen with fslview dr_stage3
> corrp_stat1 and 2 and pstat1 and 2.
> I have seen that with a threshold of 0.95 (corrp_stat1 and 2) with only 14
> subjects I have ssen activation in only two or three networks. But, with
> pstats1 and 2, with a threshold of 0.99, all my components show activation
> in the different networks. Now, I have to do the clustering, but I don't
> know how to do it...someone could help me, please?
> Thank you.
> Rosalia.
--
Dr. med. Cornelius J. Werner
Department of Neurology
RWTH Aachen University
Pauwelsstr. 30
52074 Aachen
Germany
|