Dear Sara
Thanks for sending me your DCMs. The problem you experienced was due to your resting state (CSD) DCMs being mis-specified. For resting state, the B-matrix (modulations by experimental conditions) should be empty, but specified in a particular way:
DCM.b = zeros(n,n,0);
Where n is the number of brain regions. Whereas you had:
DCM.b = zeros(n,n);
Which caused it to crash. I am wondering how this was mis-specified, as I think we resolved this in SPM quite a while ago. Please can I check, did you use existing scripts to make your DCMs, or the GUI?
One further tip, that your data has brought to my attention. When estimating your DCMs in the GUI, each DCM from each session should be treated as a separate "subject". For convenience, I recommend you click 'Select DCMs', then 'Per Model', then "New: Model" and select all your DCMs from all subjects and sessions together. (This is because the option to have multiple models per subject was intended for when you have different models of the same session, rather than different sessions. I'll update the GUI to clarify this.)
> Each subject underwent three different tDCS sessions while at rest.
> do you think I should estimate the parameters for each tDCS session separately creating one GCM for each stimulation type or can I just get one GCM for all data and then compare the sessions only within the second-level PEB framework?
Go for one GCM with all models together. Then, run a PEB analysis with regressors for the differences between your tDCS sessions.
Let me know if anything else is unclear,
Peter
-----Original Message-----
From: SPM (Statistical Parametric Mapping) <[log in to unmask]> On Behalf Of Sara Calzolari
Sent: 21 March 2019 17:35
To: [log in to unmask]
Subject: [SPM] Error during DCM estimation for PEB
Dear experts,
I am trying to estimate the DCM parameters and create the group GCM file for PEB, but I get the following error during estimation:
Failed 'DCM estimation'
Index exceeds matrix dimensions.
In file "/projects/pbic1036/spm12_new/spm_dcm_peb.m" (v7476), function "spm_dcm_peb" at line 193.
In file "/projects/pbic1036/spm12_new/spm_dcm_peb_fit.m" (v7267), function "spm_dcm_peb_fit" at line 139.
In file "/projects/pbic1036/spm12_new/config/spm_cfg_dcm_est.m" (v7479), function "spm_run_dcm_est" at line 368.
Line 193 of function "spm_dcm_peb" is:
PE = pE{i}(q);
in a section of the script called "deal with rank deficient priors"
Does anyone know how to fix this?
Few details on my DCM model: it's resting-state data so my DCM model only has a specified A matrix (not fully connected because we left out a couple of connections that do not make sense anatomically), with B being a 4x4 zero matrix (I have 4 regions) and C being a 4x1 zero matrix.
Thank you!
Best,
Sara
