Dear Wei,
It indeed looks like there is inconsistency between DCM.A and DCM.Ep
in your files. However, the true priors in DCM.M.pE and the posteriors
are consistent in the sense that you get posterior estimates different
from the prior (-32) only where the prior was set to 0. If the
structure specified in DCM.M.pE is what you intended (i.e -32 where
you put zero in DCM.A/DCM.B and 0 where you put 1) then DCM results
are valid. I don't know why there is an inconsistency with DCM.A. How
did you specify the DCMs in the first place? What did you answer to
'Use previous posteriors?' and 'Use previous priors?'.
Vladimir
On Tue, Jan 29, 2013 at 10:11 AM, Penny, William <[log in to unmask]> wrote:
> Dear Wei,
>
> As far as I remember the DCM.A, DCM.B and DCM.C fields just contain a description of the
> structure of the network - matrices of 1's and 0's telling SPM what is connected to what.
>
> Naturally, this is fixed and will not be updated.
>
> In contrast, DCM.Ep contains the actual parameter estimates - these should be different between models.
>
> Best, Will.
>
>> -----Original Message-----
>> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
>> On Behalf Of Wei He
>> Sent: 29 January 2013 03:15
>> To: [log in to unmask]
>> Subject: [SPM] DCM.mat file
>>
>> Dear all,
>>
>> Just realized some wired things happened to my DCM output files
>> (attached one inconsistent output here):
>>
>> I’ve defined three different model structures to evaluate my ERF data.
>> However, after inverting, the DCM.Ep structures were all the same
>> across the three models (the parameters inside were differed in some
>> ways), while the DCM.A, DCM.B, and DCM.C still kept the original
>> settings. Does this mean the DCM failed to recognize the differences
>> in model structures given some unknown reason? This happened to all my
>> data, is there any way I could fix the problem without re-run the whole
>> processing?
>>
>> Thanks very much in advance.
>> Wei
>
|