Hi Anastasia and Gabor,
thanks for giving me the opportunity to clarify the demeaning issue...
Generally, if you want to add a "nuisance" covariate (such as age), you can do it according to the two models we've discussed with Gabor (depending on whether you think there might be an interaction between this covariate and group).
With randomise, you don't need to demean the main EVs of interest, only the nuisance covariate (and around the mean of each group if you expect an interaction and entered the age for each group in one column each).
Now, if you do demean each column including the EVs of interest
(so instead of design.mat:
1 0 -3
1 0 1
0 1 -2
0 1 4
design.con:
1 -1 0
-1 1 0
you're going for design.mat:
1 -3
1 1
-1 -2
-1 4
and design.con:
1 0
-1 0)
then yes, you'll need to add the -D option in randomise.
Hope this helps,
Gwenaelle
--- En date de : Mer 10.3.10, Anastasia A Ford <[log in to unmask]> a écrit :
> De: Anastasia A Ford <[log in to unmask]>
> Objet: Re: [FSL] Re : [FSL] VBM design
> À: [log in to unmask]
> Date: Mercredi 10 mars 2010, 15h10
> Hi Gwen,
>
> I had a similar question regarding VBM and demeaning. When
> do we need to demean the
> age, or any other behavioral variable? Also, if we do
> demean, then we are running VBM we
> do not need to use -D option, is that correct?
>
> Sincerely,
>
> Anastasia
>
|