Dear John and SPMers,
Could you PLEASE COMMENT on my VBM study design (SPM2), because I am not
sure if I really detected the effects that I was looking for. I studied
already the archive but didn't understand everything.
I want to look for the age-effect using a regression analysis with
modulated GM images, adjusting for brain size, i.e. using the GM volumes
as a confound. In effect I intend to study region-specific GM volume
changes that correlate with age, corrected by the individual total GM
volume.
1. PET model "single subject covariates only"
2. 1 covariate: age ("no centering around grand mean"...?)
3. 1 nuissance variable: gender ("no centering around grand mean")
(should I better use the GM volumes here as additional nuissance
variable? With or without centering around mean? In consequence no
global calculation to mean voxel value and no global normalisation?
4. Threshold masking: relative 0.5 (?)
5. no overall scaling to grand mean (would it make any difference for the
analysis?)
6. global normalisation: proportional scaling global mean to 1 (1?)
(is global normalisation nessecary? if, then better proportional
scaling or AnCova?)
7. global calculation: user specified (GM volumes)
Thanks in advance for your comments,
Samuel
|