 |
 |
Dear Hoaran,
I think it is always worthwhile examining convergence issues in more detail before taking the results to
a second stage of analysis (ie BMA).
The two options are (i) increasing the number of iterations yet further or (ii) changing the initialisation point.
Re (ii) it is possible to call the optimiser (spm_nlsi_GN.m) with the field .P set
in the data structure M (ie M.P).
By default this is set to the mean of the prior distribution. (ie M.P=M.pE)
Possible ways to set this include (a) using the posterior mean parameter values obtained
by fitting the model to a different subjects
data(ie M.P = DCM.Ep where the DCM is from another subject) or by initialising some
of the parameters to those values found from fitting a similar
model.
Also, in DCM for ERP a common issue that gives rise to possible fitting problems is
the timing of the ERP components that you are trying to
explain. If you have very late components it may be a good idea to delay the input
that cortical regions receive.
Best wishes,
Will.
From:
SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of
??
Sent: 14 December 2011 11:44
To: [log in to unmask]
Subject: Re: [SPM] DCM | model convergence
Dear Burkhard,
Would you please tell me
the relationship between convergence and BMA? I confronted some problems and can't do BMA successfully. Really thank you very much！
Best regards!
--
Haoran
At 2011-12-13 16:34:50,"Burkhard Maess" <[log in to unmask]> wrote:>Dear Robert and Haoran,
>
>I would rather disagree - a model which has not reached convergence
>should not enter further analysis. With some luck the free energy value
>is low and the model is ignored by BMA, but I would not bet on this.
>
>all the best,
>Burkhard
>
>
>飞鸟 wrote:
>> Dear Robert,
>> Sometimes it will not converge within 64 iterations( or as you set
>> 128 iterations), and I think you can use Ep.A B C for further analysis.
>> Best regards.
>> --
>> Haoran LI (MS)
>>
>> At 2011-12-13 03:30:51,"Robert Schulz" <[log in to unmask]> wrote:
>> >Dear DCMers,
>> >
>> >already posted this question some time ago, I haven't received an answer yet. So please excuse if I ask again: In motor system DCMs with 6 nodes and grip left / right as modulatory influence, I don't get "convergence" after 128 iterations. What happened here. Is Ep. A B C though correct for further analysis?
>> >
>> >Thank you a lot,
>> >Kind regards to you
>> >
>> >Robert
>------------------------------------------------------------
>Dr. Burkhard Maess
>Max Planck Institute for Human Cognitive and Brain Sciences
>Stephanstr. 1a, P.O. Box 500355, D-04303 Leipzig
>Aussenstelle Bennewitz, phone/fax: +49(3425)887525-26/-11
>mail: maess 'at' cbs.mpg.de, http://www.cbs.mpg.de
>------------------------------------------------------------
>Please do not attach files which may contain executable code,
>e.g. .rtf is preferred over .doc(x)