At 19:54 30/06/2005, Elif Ozdemir wrote:
>On Thu, 30 Jun 2005 13:45:14 +0100, Christian Matthiae
><[log in to unmask]> wrote:
>
> >Hi Elif,
> >i am not an expert, so i am not completely sure about this.
> >But i think there is no problem, as you dont get an error, but only a
> >warning. Every DCM matrix contains not specified connections, which can not
> >be estimated. During estimation these connections get the value Nan, which
> >is indicated by the displayed warning. So i think your estimation was
> >finished correctly. The reason that you get this warning after the first
> >step is, that iteration stops, if the changes from one step to the next are
> >smaller than a threshold (1e-6 i think). This condition is met in your
> >analysis after the first step (dp: 1.139079e-21). This is a little
> >astonishing for me, but perhaps could be due to a very simple model?
> >If i am wrong here one of the spm/dcm should please correct me.
> >Hope this helps,
> >Christian
> >
> >
> >On Thu, 30 Jun 2005 01:26:49 +0100, Elif Ozdemir <[log in to unmask]
> >HARVARD.EDU> wrote:
> >
> >>Dear all,
> >>
> >>I am conducting a DCM analysis on a dataset that we acquired from a
>subject
> >>using a sparse temporal sampling method (TR=15, TA=1.75 sc). I have
>defined
> >>three VOIs for two (out of four) conditions. When I do the estimation, I
> >>get the error message below after the first iteration. I have read the
> >>previous posts about this error and have the latest DCM updates.
> >>
> >>After getting the error message, I am still able review the results. The
> >>histogram is drawn only for B probabilities, not for B modulatory effects.
> >>In addition, even though the VOIs are circled, there are no numbers.
> >>
> >>I have replicated the results of the attention data from Dr.Penny's
>website
> >>without these error messages.
> >>
> >>Any ideas how to fix this? Any input/help is greatly appreciated!
> >>Thanks!
> >>
> >>Elif
> >>
> >>"E-Step(-): 1 F: -1.043264e+03 dp: 1.139079e-21
> >>Warning: Returning NaN for out of range arguments
> >>> In /usr/local/spm2/spm_Ncdf.m at line 76
> >> In /usr/local/spm2/spm_dcm_estimate.m at line 72
> >> In /usr/local/spm2/spm_dcm_ui.m at line 312"
>
>Hi Christian,
>
>Thanks for the message. I have specified several DCMs with conditions
>ranging from two to 6, VOIs from 2 to 6 and intrinsic conditions from 2 to
>12. None of them worked (at least completely). A and B probabilities are
>there. But no numbers on the regions. There are dashed lines instead of
>solid ones with red dots around the VOIs.
>
>In the last DCM model, I had 8 conditions in the design matrix, took two of
>those conditions to the DCM, had 6 VOIs and 12 intrinsic connections. For
>the condition one and two, I had same number of connections (I think 6).
>The NaN error message still comes up after the first iteration.
>
>Any ideas as to how I could make these dashed lines solid and see the
>corresponding numbers? Loading DCM.A mat gives me only -1s and 0s.
>
>Any help is very much appreciated!!!
>
>Elif
Dear Elif,
is it possible that you forgot to specify driving inputs for your
model? Without such driving inputs, there will be no dynamics
whatsoever, and the model will converge at the first iteration.
Best wishes
Klaas
|