Dear Veda,
there is perfect multicollinearity between some of your regressors, so
after serial orthogonalisation, you are left with 0 for the last
parametric modulation.
You can see this with:
load level1_multivarsub01_hrf2.mat
X=[matlabbatch{1}.spm.stats.fmri_spec.sess(1).cond(1).pmod.param];
X(:,4)+X(:,6)-X(:,1) % constant
rank([ones(size(X,1),1) X]) % 6 and not 7
If you are testing for your main effect of stim (ie contrast [1 0 ...]),
this is not a problem but you could simplify your design by not
including the last parametric modulator altogether.
Best regards,
Guillaume.
On 08/10/13 08:19, Veda wrote:
> Dear SPM experts,
>
> I included six parameters in the parametric modulations. However, the
> result indicated that one parameter (i.e., maxChunkSize) was not
> unique. I don't understand why this could be a problem because the
> parameters in the parametric modulations were covariates to my
> experimental conditions. They were not supposed to be unique. I
> attached the script for the first level analysis and the resulting
> design matrix. I appreciate that someone could point out to me what
> is wrong. Thanks!
>
> Sincerely, Veda
>
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|