Dear Cyril
It is difficult to say in general anything about this. If for instance
there is no motion at all, then the mask defined by the effects of
interest will be very similar in the two cases. In other cases the
difference can be huge.
best
Torben
P.S. remind me to get back on the P.S. some other day :-)
Torben E. Lund
Danish Research Centre for MR
Copenhagen University Hospital
Kettegaard Allé 30
2650 Hvidovre
Denmark
email: [log in to unmask]
webpage: http://www.drcmr.dk
On 17 Feb 2005, at 15:14, Cyril PERNET wrote:
> Dear Torben,
>
> Thank you very much for your reply (two days ago)
> I created a matrix with 4 sessions and 9 regressors by sessions (3
> conditions convolved by the hrf + 6 motion parameters) and associated
> the data. Then, I ran the instructions you kindly send to me
> (SPM.xX.iG=4:9,13:18,22:27,31:36 and SPM.xX.iC(SPM.xX.iG)=[] ) and
> estimated the matrix.
>
> I tested different contrasts but I observed very few differences in
> comparison with the classical way: the increase of p values at the
> voxel-level was around .001 or .002. The only major difference is at
> the
> set-level. The number of expected clusters was respectively 11 and 10
> for motion parameters as regressors of interest vs of no interest (it
> is
> similar) but the probability to find this number or cluster or more
> increased from .222 to .331
>
> Does this result correspond to what you usually have? or maybe I've
> done
> something wrong!
>
> Thank you
> Cyril
>
> PS: I also have some difficulties to understand the subtle difference
> between 'of interest' and 'of no interest' when thinking at the
> estimation of the model. I'm not an expert in mathematics, can someone
> explain me the difference in simple words? What I suppose is that in
> the
> case of regressors of interest, for Y=f(Xi + ...Xk) + Ei, the motion
> parameters are included as observations (Xi) whereas in the case of
> regressors of no interest, motion parameters are included in the E?
|