Hi Dimitrios
> what is the best way to control for both intracranial volume and global GM effects? Is it permissible to include ICV as a covariate and then use ANCOVA for global GM effects? Or should one use ICV as user defined global values for ANCOVA. The concern is the potential high correlation between ICV and global GM and the resulting co-linearity in the model.Thanks,
Do ICV and total GM actually show strong correspondence in your data?
This isn't always the case...in cases of dementia or age-related
cortical volume loss, ICV stays relatively stable but GM will decline
(e.g. Good et al. 2001, Fig. 3). In general I would think that one or
the other would be sufficient, depending on the population you are
looking at. Alternatively you could orthogonalize them before
entering in your model.
Or, if you are just entering them to control for global effects, maybe
colinearity isn't such a problem, as your main concern is that
variance attributable to either is assigned to one of those
regressors, and it's therefore not an issue if it's assigned
'incorrectly'.
Hope this helps,
Jonathan
|