Dear Matteo,
I assume when you refer to fits, you mean the R^2 ? As you say, stochastic DCM will generally do a good job of fitting the signal, making sanity checking a little harder. I would suggest you base your judgement on the size of the parameter estimates. The larger they are, the more they're contributing to the model. Karl wrote a tool for this - spm_dcm_fmri_check() , which is included in SPM. The size of the parameters are at the bottom left, and will be shown in red if they're all negligible.
Best,
Peter.
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Matteo Tonietto
Sent: 26 August 2014 22:21
To: [log in to unmask]
Subject: [SPM] stochastic DCM evaluation
Dear all,
I just started using stochastic DCM and compared to deterministic DCM I am obtaining better fits of the data.
However, I am obtaining good fits with all the models I have tested and I am wondering if for some of them the stochastic part is explaining all the variance in the data and the model structure is essentially negligible.
How do I understand if stochastic DCM fitted the data correctly?
Thank you for your help
Matteo Tonietto, PhD Student
Department of Information Engineering
University of Padova
|