Hello,
> I am doing aVBM group study using SPM. To account for brain size I am doing
> a global correction as follows
>
> 1. Calculate volumes for gray matter, white matter and CSF using
> spm_get_volumes.
> 2. Add these up and feed them as globals in the design specification.
> 3. Account for the globals by Proportional scaling or adding as a covariate.
>
> I notice that when I do this the t-scores increase quite a bit and I get
> blobs that survive whole brain correction (more of them survive if I do
> scaling instead of adding as a covariate). But if I don't include globals at
> all and don't perform any global normalization the t-scores are quite less
> and nothing survives whole brain correction.
>
> Is it normal to see this difference? Which is a better method of global
> normalization?
It makes sense that you might see different results depending on
whether you include global values (either TIV or total gray matter).
In general, if you are able to account for more variance in your data
(which TIV will likely do), you may increase sensitivity to effects
(and thus see more significant results). However, if the
covariate/scaling is correlated with some other measure of interest,
the interpretation can get tricky.
You may also be able to find some more comments in older posts—for
example, some further discussion of this issue can be found in this
message (from yesterday):
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;19064809.1211
Hope this helps!
Best regards,
Jonathan
--
Jonathan Peelle, PhD
Department of Otolaryngology
Washington University in St. Louis
Office: (314) 362-9044
http://peellelab.org || http://jonathanpeelle.net
|