Hello,
I am interested in testing for positive and negative correlations between
age and activation, while regressing out the effects of RT. I am
specifically confused about the difference between demeaning an EV versus
orthogonalizing that EV wrt to other EVs.
Model #1
EV1=mean
EV2=age (not demeaned)
EV3=RT (not demeaned)
Orthogonalize EV1 wrt EV2
Orthogonalize EV3 wrt EV2
C1= [0 1 0]
C2= [0 -1 0]
My interpretation is that C1 will give voxels showing a positive correlation
between age and activation (and C2 a negative correlation), with the
variance associated with both RT and the global signal regressed out. Is
this interpretation correct? If so, how is this different from Model #2
(below).
Model #2
EV1=mean
EV2=age (demeaned)
EV3=RT (demeaned)
C1= [0 1 0]
C2= [0 -1 0]
Thanks!
Libby
|