Hi,
The orthogonalization option is only necessary if you want to
put contrasts on the motion parameters (i.e. get statistics about
the motion parameters themselves). If you just want them to
remove as much variation as they can from the timeseries
then using them as confounds is all you need to do. Hence
we thought that it was simpler, easier and exactly the same
result for the vast majority of situations if we removed the
orthogonalization button.
If you check the radio box for "add motion parameters to
the model" then it will take these from the mcflirt run in the
prestats. If you do not run mcflirt in the prestats then you
will need to get your motion regressors some other way
and enter them as confounds yourself.
So if you just check the button then you will get what you
want, with the motion regressors added automatically and
they will soak up as much variation as possible (the same
as if you orthogonalize).
All the best,
Mark
On 26 Feb 2009, at 16:41, David Paulsen wrote:
> I'm running a study with children, who tend to exhibit a little more
> motion
> than adults typically do. I'm finding lateral ventricle activations
> and
> would like to consider orthogonalizing my EVs with respect to the
> motion EVs
> generated from mcflirt. I was told that this option is no longer
> possible
> with the current version of Feat. I have a couple of questions.
>
> 1) Is there any orthogonalization between motion and other EVs
> taking place
> in the current version of Feat by default?
>
> 1a) If not, is it possible?
> 1b) What is the rationale for excluding this option?
>
> 2) If prestats are run before first level stats, is it true that
> checking
> the radio box "add motion parameters to model" at the first does not
> direct
> Feat to grab the mcstats from the prestats directories?
>
> David
>
|