This looks like you just want a linear regression where
the voxels and time axes are swapped, meaning you
would find a time series of TR coefficients of one 3d+time
dataset that have a minimum least squares fit to the other
3d+time dataset. The beta weights are just over time,
rather than over space. They would scale one dataset
(the model, say) to fit the other - there would not be betas
for both datasets.
If so, that is not hard to do. I have given commands to do
it in the past.
Is that what you want?
- rick
