> I am using VBM with SPM5 to compare 2 groups? non-functional MRI
> scans. I found a 2003 study which used SPM 99 and used ANCOVA to test
> for differences between patients and controls in both density and
> volume of grey and white matter. The authors used the integrals of GM
> and WM, respectively, to control for global differences in the amount
> of GM and WM to allow them to test for regional differences. The
> authors chose this course of action based on SPM mailing list
> feedback on Oct 25, 2000 (Subject heading, "Scaling of global effects
> and VBM").
> I would also like to control for global differences to test for
> regional differences. However, when I search for ANCOVA in the SPM5
> manual, the manual states that ANCOVA is only for use with PET data.
> Does the latest version of SPM/VBM suggest a different approach? How
> should I proceed?
It depends on the kind of differences you would like to identify. If you want
absolute differences, then don't include any globals. If you want only
differences that can not be explained by changes in total brain volume, GM
volume, intra-cranial volume etc, then you can use an ANCOVA model.
Alternatively, you may want to analyse each region in terms of its proportion
to the total GM volume (or some other measure), in which case you could use a
proportional scaling model.
I think most people use an ANCOVA model. For dementia studies, TIC is usually
modelled as a confound. Other studies may covary out different measures. A
quick search through the SPM mailing list for the words "global and VBM"
should give you plenty of hits.