Re: scripts -- January 17, 2007 - 05:42

I would suggest adding the following line just before the Monte Carlo
simulations:

fix('aic', 'diff', fixed=1)

Optimising the global model (diffusion tensor + all model-free models)
is computationally expensive and hence optimising solely the
model-free models during the Monte Carlo simulations would save you a
lot of time.  The only problem with doing this is that the parameter
errors are slightly underestimated.

Actually I would suggest axing the Monte Carlo simulations and
rerunning the model-free optimisation, model elimination, model
selection, and global optimisation many times.  The diffusion tensor
will change across the iterations.  Once you have obtained convergence
(exactly equal chi-squared values, identical model-free models, and
equal parameter values), then and only then would I recommend running
Monte Carlo simulations.  If you feel adventurous, you could have a
try of the new model-free optimisation protocol embedded in the
'full_analysis.py' sample script.  If you use the new 1.2.10 relax
version, the 'full_analysis.py' script will tell you when convergence
has occurred.

Edward


On 1/16/07, Hongyan Li <hylichem@xxxxxxxxxxxx> wrote:
> Hi, everyone,
> I used a mf-multimodel.py (similar to the sample script) to generate m0 to m9
> models and then used modsel.py to select a suitable model for each spin and
> then minimised all parameters using selected models. I would also like to do
> Monte Carlo simulations at the last stage, but combining these together, it
> will take too long to run. Any suggestion how to improve it. Ths script I used
> is like this:
> # Script for model-free model selection.
>
> # Nuclei type
> nuclei('N')
>
> # Set the run names.
> runs = ['m0', 'm1', 'm2', 'm3', 'm4', 'm5', 'm9']
>
> # Loop over the run names.
> for name in runs:
>     print "\n\n# " + name + " #"
>
>     # Create the run.
>     run.create(name, 'mf')
>
>     # Reload precalculated results from the file 'm1/results', etc.
>     results.read(run=name, file='results', dir=name)
>
> # Model elimination.
> eliminate()
>
> # Model selection.
> run.create('aic', 'mf')
> model_selection('AIC', 'aic')
>
> # Minimise all parameters.
> fix('aic', 'all', fixed=0)
> minimise('newton', run='aic')
>
> # Monte Carlo Simulations
> monte_carlo.setup('aic', number=500)
> monte_carlo.create_data('aic')
> monte_carlo.initial_values('aic')
> minimise('newton', run='aic')
> eliminate(run='aic')
> monte_carlo.error_analysis('aic')
>
>
>
> # Write the results.
> state.save('save', force=1)
> results.write(run='aic', file='results', force=1)
>
> With best regards,
>
> Hongyan
>
> Dr. Hongyan Li
> Department of Chemistry
> The University of Hong Kong
> Pokfulam Road
> Hong Kong
>
>
> _______________________________________________
> relax (http://nmr-relax.com)
>
> This is the relax-users mailing list
> relax-users@xxxxxxx
>
> To unsubscribe from this list, get a password
> reminder, or change your subscription options,
> visit the list information page at
> https://mail.gna.org/listinfo/relax-users