Dear Christina,
Many thanks for reporting this and for your detailed analysis. Your proposed solution is nearly right (it wouldn’t allow for multiple parametric regressors). I have now fixed this for future SPM releases and I attach the fixed version of spm_dcm_U.m. Please let us know if you have any more difficulty with it.
Best,
Peter
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Christina Regenbogen
Sent: 30 September 2015 07:59
To: [log in to unmask]
Subject: [SPM] DCM parametric modulator 2nd order (linear and quadratic)
Dear SPM Mailing list users,
we set up a design matrix with two task-regressors, the second one including a parametric modulator, modelling linear and quadratic changes (2nd order).
SPM.U thus has two entries, where
SPM.U(1) = task1, and SPM.U(2) = task1 pmod1 pmod2
We encounter problems setting up a DCM in which we include all 4 inputs - not during the original model specification with the GUI but when adjusting the DCM for a new participant in DCM input specification
matlabbatch{1}.spm.dcm.fmri.inputs.val = {1,[1 1 1]};
The model estimation gets errored out because of lines 66-69 in spm_dcm_U.m (error(['More parametric modulations specified than exist ')
For i=2 (second regressor)
mo = find(inputs{i}) is 1 2 3 and its length is therefore 3.
Therefore, length(mo)-1 (=2) is in fact greater than length Sess.U(i).P (=1) and the error is true.
We can, however, run the model if we replace length(Sess.U(i).P) with Sess.U(i).P.h (which returns 2 in the case of 2 pmods) and the respective U structure seems indeed appropriate.
Has anyone else encountered this and does this signal that it is actually discouraged to utilize DCM for a higher than 1st ord pmod modellation design?
We would appreciate your input on this.
Best regards and thank you,
Christina
Christina Regenbogen | Postdoctoral researcher
Dept. of Clinical Neuroscience | Karolinska Institutet
[log in to unmask]<mailto:[log in to unmask]>
