Hi Gordon,

I'm afraid I'm a bit confused about exactly what you're doing... are
you extracting a beta for each subject and trying to use these
directly? Or forming a "con" value for each subject from betas and
using those? Or extracting data from each subject's con image? And
when you say "extracting for a region" do you mean a particular voxel,
or averaging over voxels within a cluster/ROI, or using SPM's VOI
extraction (which uses the first eigenvariate rather than the mean),
or something else?

If you extract a particular voxel from each subject's con image and do
a two-sample t-test of these (with the same choice of equal or unequal
variance that you specified in SPM) then you should be able to
reproduce that voxel's t-value. I've done this a few times.

As a wild guess, I imagine that if you had quite rough data, then
averaging over a largish region of significant *voxels* might lead to
a non-significant result for the average. As an artificial example, if
you consider two neighbouring voxels, both of which are larger in
group B than in group A, but with the first voxel being negative (i.e.
more negative for A), and the second voxel being positive (more
positive for B), I would guess that their average could show less of a
significant difference. However, I haven't thought about this
carefully or tried it out though...

If this hasn't answered your question, it would help if you posted to
the list more details, including my questions above, and also things
like: what resel sizes your whole-brain analysis reports, and how
large the cluster of significance is, etc.

Best,
Ged


On 05/11/2007, Waiter, Dr Gordon D. <[log in to unmask]> wrote:
> Having performed an RFX 2-sample t-test on two groups (taking the con
> images from condition X vs Rest for each participant), I then decided to
> extract the beta values for one of the regions of significant difference
> in each participant and to my surprise I found that there was no
> apparent difference when looking at the data this way.
>
> Could someone possibly shed some light on why the difference in
> extracted betas does not show a significant difference when the whole
> brain analysis does?
>
> Thanks in advance,
>
> Gordon
>
> Gordon D. Waiter PhD CSci MIPEM CPhys MInstP
> MRI Physicist
> Department of Radiology
> University of Aberdeen
> Research MRI Centre
> Lilian Sutton Building
> Foresterhill
> Aberdeen
> AB25 2ZD
>
> Tel: +44 (0)1224 559725
> Fax: +44 (0)1224 559718
>
> [log in to unmask]
> www.abdn.ac.uk/radiology
> www.abdn.ac.uk/ims/imaging/
>