Hi Lora,
If you have your log messages saved to file (using the --log or --tee
options), then you'll see the model selection messages close to the
end. You can also load the final relax state file and run the
diffusion_tensor.display user function to see something like:
"""
relax> diffusion_tensor.display()
______________________________
Diffusion type spheroid
tm (s) 9.08889e-09
Diso (rad/s) 1.83374e+07
Da (rad/s) 7.13232e+06
Dpar (rad/s) 2.30923e+07
Dper (rad/s) 1.596e+07
Dratio 1.44689
theta (rad) 0.96817
phi (rad) 2.47883
Fixed flag True
______________________________
"""
As for the model-free model for each residue, this can be saved to a
file via the value.set user function where the param argument is set
to 'model'. This is deliberately not performed by default as a
concious design decision. The reason is because historically there
has been too much emphasis on the specific model selected, whereas the
emphasis should be rather on the dynamics. The functional
significance of the model-free motions is revealed by the dynamics
itself rather than which model was selected. As there is a lot of
nesting in the model-free models, and just because a model with no Rex
has been selected does not mean that there is no Rex, just that no
statistically significant Rex can be found with the current relaxation
data. And sometimes two nested models can be equally good via
standard parsimony-based statistics (AIC for example) and which one is
chosen will be close to arbitrary. Careful post analysis,
meticulously checking all data, will reveal a number of these cases.
The more complex model will be noisier and its additional parameter
will be on the baseline of statistical significance for that
parameter, below which that parameter is invisible (it could still
exist). The simpler model is less noisy and the additional parameter
is then considered part of the invisible region - whether or not it
exists can never be determined. All model-free parameters have such
statistical invisibility bands within the parameter space, where 2
times the lower error bar will hit the zero-intercept. This is why
de-emphasising models is a good idea. If you were to remeasure the
relaxation data with higher precision, the dynamics picture will
simply become less noisy and you'll be able to see more into the
invisible parameter regions. However the model selection picture will
change radically. Anyway, just run the value.set user function to
obtain the models selected (or look inside the XML state file).
Regards,
Edward
On 13 April 2015 at 23:01, Lora Picton <lkpicton@xxxxxxxxxxxx> wrote:
Hi,
Is there a file in the final results that says which model that each
residue fits best to, and which of the overall models of the molecule
(sphere, ellipsoid etc) was the best fit?
Thanks for your help.
Sincerely,
Lora Picton
_______________________________________________
relax (http://www.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
Re: question about final files generated