Dear Filip
I hope you don't mind me CC'ing the SPM mailing list so others can benefit.
> If I run the second level analysis on 10 Subjects I get the error that
PEB.F = NaN
> Warning: Matrix is singular, close to singular or badly scaled. Results
may be inaccurate.
RCOND = NaN.
> In expm>PadeApproximantOfDegree at 123
In expm at 39
In spm_dx at 165
In spm_dcm_peb at 434
In PEB_DCM_second_level_41Subj_sham at 199
> VL Iteration 24 : F = NaN dF: 193.6026 [-4.25]
> VL Iteration 24 : F = NaN dF: 193.6026 [-4.25]
> But if I do GCM estimation and second level PEB for 3 Subjects (testes
for all Subjects (1,2,3)(3,4,5)(5,6,7)(7,8,9),(8,9,10) every Subject
seems fine and I have no Problem with PEB.F
> some values PEB.F = 528.5658 ; PEB.F = 639.3666 ; PEB.F-15.620 ;
PEB.F=102.1413
> I've tested some more constellation and even the analysis of 4 or 5
Subjects give me PEB.F= NaN
Your PEB estimation has failed, which suggests the priors in the PEB model aren't right for your DCMs. Usually, this means that something has gone wrong with the DCMs - which may just be there are no consistent effects across subjects. I suggest you load your GCM array into the workspace and run:
spm_dcm_fmri_check(GCM);
This will provide you with a tool to investigate the explained variance and A-matrix parameters of each subject's model. If you have low explained variance (< 10%) for many subjects or very small A-matrix parameters (< 0.1) it suggests there has been an issue with model fitting.
Let us know if you have any further questions in trying to diagnose this.
Best
Peter
|