Dear Peter and DCM experts,
I have some questions regarding the Two-state DCM, and I would appreciate if you answer them. I am using the SPM8, DCM10.
1. The model that I am using to estimate the parameters consist of 3 regions which are connected in a feed forward way to each other and so for the A matrix I have 3 zero components. after running the estimation I receive the value -32 for the Posterior of the A components which were zero! and as I checked the function "spm_dcm_fmri_priors" I found that this is defined for the priors expectations "pE.A = A*32 - 32" for matrix A, which seems to be reason I got this value! Do you what is problem and how can I fix it? I also tested with SPM12 which they have used the same priors for matrix A.
2. For a 3 region single state DCM I have a 3*3 A Matrix, I want to know for two state DCM I again should define a 3*3 A matrix or 6*6 according to main paper by Marreiros, et. al?
3. In discussion part of the Marreiros, et. al, paper (the sixth paragraph) it is mentioned that one can use instead of linear state, a nonlinear function of states (sigmoidal functions which were introduced by Wilson & Cowan 1973). I wan to know if you or any DCM expert specially the writers of paper know if anyone has used these kind of functions for their state in fMRI? I would appreciate if you introduce me some papers (and also the codes if they are available).
4. My last question is about the SPM software. As the researcher in my institute use SPM8 (DCM10) for their researches and believe that DCM10 give a better result than SPM12. I want to know what you think about this?
Best regards,
Sadjad
|