Hi Tyler,
My feeling is that it would be a bit hit-and-miss whether 2D
registration would help. One would certainly need to work
hard to get the subjects positioned straight in the scanner.
If a nod isn't just a sagittal motion, all bets are off, as
the motion goes out of plane.
---
In any case, I would not expect the extra regressors to be
close enough to be a worry in terms of collinearity. That
would be an equal concern for the normal 6 parameters, yet
it is very rare to see any warnings.
But even if they were close to collinear, one could probably
ignore it. It doesn't matter which of the motion regressors
eats up more variance, it just matters that motion is accounted
for in the regression. As long as the regressors of interest
are not similar to the motion parameters, it should be okay.
- rick