Hi
I have
two questions concerning VBM analysis and covariates. My analysis is set up with
randomise as a t-contrast, between smokers and non-smokers, with
covariates for gender, age, PTSD (high cormidity with smokers in this sample),
and total intracranial volume (GM + WM +CSF).
The
first question is whether the the use of total intracranial volume
(TIV) is an appropriate covariate ? It was suggested by a collaborator on
this study. I understand that in FSLVBM voxels are modulated by the
Jacobian derived from the normalization step. Does this modulation address
the issue of TIV or am I confusing two issues that have nothing to do with
another. Also, a minor point is that Good (2001) describes 'modulation' as
multiplication by the determinant of the Jacobian whereas the FSLVBM website
describes dividing by the Jacobian. Any clarification would be
helpful.
My
second question is how to introduce a covariate for number of pack years of
smoking. For obvious reasons the number of pack years (in smokers) is highly
correlated with the age of the subject. How can I use pack years as an
explanatory variable (covariate) when it is strongly linked to age, and that age
has a robust relationship with loss of gray matter volume ? Would I use
both pack years and age as covariates and expect that the GLM will work this out
because pack years is zero (or at least very low) in the non-smoker group
?
Thanks
raj