Hi,
On 10/20/06, Daniel Perez <daniel.perez@xxxxxxxxxxxxxxxx> wrote:
Hi Ed,
I ran calculations using full_analysis.py with two data sets. After
runing the diffusion models and getting convergence for all of them I
tryed with the model=final. I got the following error messages:
Dataset 1, model=final
Traceback (most recent call last):
[snip]
File "/home/perez/relax-1.2.6/generic_fns/eliminate.py", line 69, in
eliminate
names = param_names(self.run, i)
File "/home/perez/relax-1.2.6/specific_fns/model_free.py", line 1448,
in get_param_names
if not self.relax.data.res[self.run][j].model:
AttributeError: Element instance has no attribute 'model'
This was reported as bug #7226 (https://gna.org/bugs/?7226) in the
relax bug tracker (https://gna.org/bugs/?group=relax) by Alex Hansen.
The issue has been fixed in the repository but no version of relax has
been released since then. There are two options. Version 1.2.8 of
relax could be released in the next few days (I was planning on doing
this in a few weeks anyway). Or you could bypass the release cycle by
directly downloading the relax source code from the repository. That
way you don't have to wait for me to make the new release. To do
this, you need to install the program 'subversion'. Then you type:
$ svn co svn://svn.gna.org/svn/relax/1.2 relax
to download the stable 1.2 line. The files will be placed in the
directory called 'relax'. If this doesn't work (firewall issues), you
can type:
$ svn co http://svn.gna.org/svn/relax/1.2 relax
If you would like to wait though, I can create version 1.2.8 for you.
Yeah, I gave up with windows... I have only one computer with windows
and it was toooooo slow. With the Linux network it is a piece of cake to
spread the jobs, and with parallel calculation it make sense.
About the error messages they may be related with the fact that the
local-tm was the model with the lowest chi square, I do not know.
Yep, that is the issue. You must be working with a very dynamically
interesting system if that is the AIC selected global diffusion model!
Well, knowing that this protein does not fit any diffusion tensor I used
relax to sample the reduced spectral density function. I obtained the
list containing the fitted values for J(0), J(N), J(H), and the
corresponding errors. But, in addition you provide with a list of ALL
the Montecarlo simulations. What do you suggest to do with it?
Nothing. relax provides all the results so that you can do anything
you wish with them. For instance creating a grace plot of say S2 vs.
chi-squared for all simulations. These types of distribution plots
may be interesting when bizarre results appear.
Edward
Re: error messages with model=final and Monte carlo output