[prev in list] [next in list] [prev in thread] [next in thread]
List: relax-users
Subject: SV: Palmer model selection
From: Carl Diehl <Carl.Diehl () domain ! hid>
Date: 2010-02-25 7:01:37
Message-ID: D1DBBF1D3E3030439060EFCDE32927A6C1A38B1593 () domain ! hid
[Download RAW message or body]
Hi.
Well, the mail server said that it tried sending three times before giving up. I \
believe it has something to do with the mail servers at the university.
Anyway, sofar I've used relax 1.2.XX both for fitting and modelfree-selection.
Is there a any major difference in the fitting routines between the 1.2 and the 1.3 \
line?
So far I've used a modified full_analysis.py script (in relax 1.2.14) for analysis \
which works well and is able to accurately reproduce older modelfree-fits. I haven't \
made the transition into 1.3.4, since I would have to change from runs into pipes.
My basic idea is to rerun the my analysis script, only this time using the Palmer \
model-selection instead of AIC, which I've been using sofar.
> From an old fit using Modelfree4 and Fast-MF (1field only), I know that there are \
> only small differences in S2 compared to the 2field-data fitted using relax.
I can easily redo my 2field fits using Modelfree4 and Fast-MF, although I would \
prefer to use relax. Time-wise, the full_analysis scripts seems to be much faster \
than the Fast-MF way.
I'll take a look at the repository and try to see if I can get it to work.
Regards,
Carl Diehl
_______________________________________
Från: edward.dauvergne@domain.hid [edward.dauvergne@domain.hid] för Edward \
d'Auvergne [edward@domain.hid]
Skickat: den 24 februari 2010 16:07
Till: Carl Diehl
Kopia: relax-users@domain.hid
Ämne: Re: Palmer model selection
Hi,
It's strange that your mail bounced. Did you use a different email
address? Maybe the mail server was down for a little while :S As for
your question, did you use the Newton optimisation provided by relax
(via minfx https://gna.org/projects/minfx/) or did you use relax to
control Art Palmer's Modelfree4 program? If you have data at more
than one field strength, I would highly recommend you run relax-1.3.4
(or the repository 1.3 version) with the full_analysis.py script and
then compare. This implements the new protocol I presented in:
d'Auvergne E. J., Gooley P. R. (2007). Set theory formulation of the
model-free problem and the diffusion seeded model-free paradigm. Mol.
Biosyst., 3(7), 483-494. (http://dx.doi.org/10.1039/b702202f)
and:
d'Auvergne, E. J. and Gooley, P. R. (2008). Optimisation of NMR
dynamic models II. A new methodology for the dual optimisation of the
model-free parameters and the Brownian rotational diffusion tensor. J.
Biomol. NMR, 40(2), 121-133.
(http://dx.doi.org/10.1007/s10858-007-9213-3)
The advances implemented in this script solve many, if not all, of the
problems people encounter in a model-free analysis. And I would be
willing to bet that your results will be at least slightly different
and better after running this!
As for re-implementing the Mandel et al., 1995 model selection in
relax, this depends on if you used relax or Modelfree4 as the
optimisation engine. In any case, what you can do is develop code
inside your own script to do this. If it works, then the code can be
tidied up and shifted into the core of relax for the benefit of all
other users (or it can be added as a sample script with your name at
the top, and maybe a in the end reference to help with citation
counts). The Mandel et al., 1995 paper would be the best reference to
start to implement this. And Wikipedia would also help. If you would
like to add code directly to relax to do this, that would be very much
accepted too.
A good starting point would be my old code for implementing this model
selection. I've looked back at the history and found the file needed.
It is located in version 0.2 of relax (relax was called mf back
then):
http://svn.gna.org/viewcvs/relax/
This line was last touched 7 years ago!!! Since then the file has
been deleted as it no longer was functional. Luckily having relax in
a repository means that any file from any point of time can be
accessed. Therefore it can be restored and added back to
generic_fns/model_selection.py. To make it functional again will
require a lot of work (see my previous message at
https://mail.gna.org/public/relax-users/2010-02/msg00003.html). But
the file is located at:
http://svn.gna.org/viewcvs/relax/0.2/model_selection/palmer.py?rev=298&view=markup
Actually, I found that the file was deleted at revision r697 in the
0.3 line. The most up to date version before deletion is therefore:
http://svn.gna.org/viewcvs/relax/0.3/model_selection/palmer.py?rev=696&view=markup
I hope this helps!
Cheers,
Edward
On 24 February 2010 12:48, Carl Diehl <Carl.Diehl@domain.hid> wrote:
> Hi.
>
> I got a bounce from the gna-server, so I'm sending this again.
>
> At the moment I have a bunch of data sets that have been fitted with relax-1.2.14 \
> using the AIC model selection. My supervisor wants to test the data sets using the \
> Palmer methodology instead of using AIC or something similar.
> I've used FAST Modelfree before, so for fast results I'm running that at the \
> moment. Since I'm much more used to relax, I'd prefer to use that instead.
>
> In order to add the Chi2 and F-tests, is it possible to do this within a relax \
> script or would I need to dig into the source code?
> /Carl
>
>
> ________________________________________
> Från: Edward d'Auvergne [edward.dauvergne@domain.hid]
> Skickat: den 12 februari 2010 11:34
> Till: Carl Diehl
> Kopia: relax-users@domain.hid
> Ämne: Re: Palmer model selection
>
> Hi,
>
> This model selection method used to be supported in relax. No one has
> used it in many, many years, so this code has deteriorated and appears
> not to work anymore. I think this was last used in the 0.3 relax
> versions well over 7 years ago. The deterioration might be related to
> relax now using proper STAR format parsing of Art's mfout file. The
> F-tests are only implemented in Modelfree4 and not relax, so this
> binary needs to be used as an optimisation engine in relax.
>
> At the moment I don't have the time to repair this. This requires
> minimally the extraction of the chi-squared values and F-test info
> from the modelfree4 mfout file. This code is so old though that it
> may require adding the functionality to execute Modelfree4 with the
> model pairs to produce the F-test info. Alternatively if relax is
> used as the optimisation engine, the chi-squared and F-tests could be
> added directly to relax. In any case, this won't be trivial. But if
> anyone would like to contribute this functionality, it would be most
> welcome for comparisons and historical purposes. If you require fast
> results, I would recommend looking at the FAST ModelFree
> (http://xbeams.chem.yale.edu/~loria/software.php) or Art's scripts.
>
> Regards,
>
> Edward
>
>
>
> On 12 February 2010 10:50, Carl Diehl <Carl.Diehl@domain.hid> wrote:
> > Hi.
> > Is there any version of relax with a working model selection using the Palmer \
> > method? Both in relax-1.2.15 and -1.3.4 the method don't seem to be supported?
> >
> > If not, how difficult would it be to implement it in relax?
> >
> > Regards,
> > Carl Diehl
> > Biophysical Chemistry
> > Lund University
> > _______________________________________________
> > relax (http://nmr-relax.com)
> >
> > This is the relax-users mailing list
> > relax-users@domain.hid
> >
> > 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
> >
>
> _______________________________________________
> relax (http://nmr-relax.com)
>
> This is the relax-users mailing list
> relax-users@domain.hid
>
> 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
>
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic