Hi,
This is certainly possible, as SIENAX is measuring a global change in
brain volume
while VBM is trying to localise changes at a voxel (or at least very
local) level.
Hence there may not be enough statistical power to detect local
changes but
the global effect might still be detectable since it averages over a
greater number
of estimates, reducing the standard error for the average
measurement. Remember
that when a null hypothesis test fails it does not necessary mean that
there is no difference,
it can mean either that (a) you did not have enough statistical power
to detect
a difference, and reject the null hypothesis, or (b) there was no
difference. You cannot
tell whether it is case (a) or (b). So given the result from SIENAX I
would assume it
was (a) and that only SIENAX had sufficient sensitivity to detect it.
I would also recommend trying to examine the VBM results at different
probability
thresholds to see if there is a non-significant change between the
groups which
would be consistent with the SIENAX results.
All the best,
Mark
On 24 Jun 2009, at 15:37, Eytan Raz wrote:
> Dear FSL experts,
>
> I have a coohort of patients whom I compared to a coohort of matched
> controls. I did this both with VBM (for grey matter, with SPM5) and
> with SIENAx of FSL. Interestingly, while VBM failed to detect any
> difference between the two groups, actually, sienaX showed a
> difference in NBV and NCV. How can I explain the discrepancy between
> these two results?
> Can I accept both results (sienax and VBM-SPM )of them or is there
> surely a pitfall somewhere in the analysis?
>
> Thanks, Eytan
|