At 12:57 22/03/2005, you wrote:
>Hi I have a question concerning the computation time for parameter estimation.
>
>I have now set up my first DCM model, 4 regions of interest, 5 connections.
>The design has 4 regressors, every regressor is affection every connection.
>
>The parameter estimation runs now for nearly 24 hours.
>System Pentium IV 2.5 Ghz, 2 GB Working mem. The maxmem parameter in
>SPM_defaults ist set, so that SPM2 con use up to 1 GB of system
>rescources, but
>this was only needed in the first hour of parameter estimation.
>
>The time series consists of roughly 1150 scans, voxel size 3mm, ROI
>speroids with
>6mm radius.
>
>Is this computational time needed normal, or is there something going
>wrong, unfortunately
>the matlab protocol widow is empty. (NoJava option)
>
>Thank you!
>
>
>------------------------------------------------------------
>Ferenc Acs
>Lehrstuhl Prof. Dr. M. W. Greenlee
>Institut für Psychologie
>Universität Regensburg
>93040 Regensburg
>Tel. +49 (0)941 943 3582
>Fax +49 (0)941 943 3233
>http://www.psychologie.uni-regensburg.de/Greenlee/team/Acs/acs.html
Dear Ference,
this is indeed quite a long computation time for the model you describe. I
suspect this may be a problem of convergence during parameter estimation.
Does your model converge at all and if so, how many iterations does it
need? Is it possible that you are still using the old version of
spm_nlsi_GN? With the new version, which Karl wrote a few months ago,
problems of convergence have disappeared (or are, at least, extremely rare
- personally, I have not observed any non-converging case with this
estimation scheme). The new version is now in the SPM2 Updates directory.
Best wishes,
Klaas
_____________________________________
Dr Klaas Enno Stephan
Wellcome Dept. of Imaging Neuroscience
12 Queen Square, WC1N 3BG, London, UK
phone: +44-207-8337485
fax: +44-207-8131420
web: http://www.fil.ion.ucl.ac.uk/~kstephan/
|