Dear experts,
Sorry, I have posted on the list before (a couple of months ago) regarding
these questions but have not received any response. I hope someone can
find time to help me.
I am using SPM5.
The first questions involve VBM.
I know that I should not use the cluster p-values in this case unless I use
nonstationarity correction (available in a toolbox which as yet I have not
used).
With my data I have an a priori assumption that there will be correlations in
certain brain regions. When this is the case I think it is appropriate for me to
select the cluster maxima in my a priori region and click SVC e.g. at 10mm. I
will report which clusters have received a SVC in my article. After doing a SVC
the p-values of course change and I wonder which column I should report e.g.
should it now be the cluster uncorrected p-value (or do I leave this alone
completely with VBM?), or the voxel P FWE-corr or the voxel P FDR-corr?
Or can I only use SVC if I use the nonstationarity toolbox?
If this is the case is it the uncorrected p-value I report after doing the SVC?
The last question regards fMRI.
Here it is valid to use the cluster P corrected values. I can report these in my
paper and that is fine.
When I carry out a SVC e.g. 15mm, on the maxima of a cluster (in an a priori
area) in this case which p-value should I report? Should I report the cluster P
uncorrected value (as the expected number of voxels per cluster will be
reduced)?
I take it that I should not report the cluster P corrected value as the expected
number of clusters might well be reduced below one with such a small area and
these P-values would therefore be misleading?
Can someone please clarify and answer all these questions please. I am not a
mathematician or statistician, so please keep the answers basic and
understandable. Saying that I want to make sure I report all of these things
correctly.
Thanks in advance.
Will
|