Hi,
I've now created two branches within the relax repository for you to
play with. The first is a copy of the 1.2 line and is located at
svn.gna.org/svn/relax/branches/consistency_tests_1.2/. The second is
a copy of the 1.3 line and is located at
svn.gna.org/svn/relax/branches/consistency_tests_1.3/. I've initially
used 'svn cp' to create the 5 consistency_tests.py files as described
in https://mail.gna.org/public/relax-devel/2007-07/msg00001.html
(Message-id:
<7f080ed10707090251ve1c4a8fl7f8618843e5c9459@xxxxxxxxxxxxxx>).
Would you be able to create patches for these files (in the 1.2 line
first, no need to worry about the 1.3 line yet), and then post the
individual patches as text file attachments to the mailing list?
Thanks. I will then be able to commit these patches individually,
checking them in fine detail.
Things to note in creating the patches from the code at
http://maple.rsvs.ulaval.ca/mediawiki/index.php/Relax_development
include the copyright preservation, a number of integers in
'maths_fns/consistency_tests.py' which should be floating point
numbers (just add '.0' to the end of the number), the addition of
Grace plots as an output in the 'sample_scripts/consistency_tests.py'
script to be able to create a picture similar to that on your relax
development site, maybe only allowing 'Tc' and 'tc' in the
return_data_name() function and not 'TC', and 'degrees' instead of
'degree' in the return_units() function.
One bug includes:
setattr(self.relax.data.res[self.run][index], 'csa',
float(value[0]))
setattr(self.relax.data.res[self.run][index], 'r',
float(value[1]))
+ setattr(self.relax.data.res[self.run][index],
'orientation', float(value[1]))
+ setattr(self.relax.data.res[self.run][index], 'tc',
float(value[1]))
value[1] has been used twice. I have a feeling there is another bug
somewhere where an index has been repeated a few times when it should
be different indices, but I can't find it at the moment. The
individual patches should help. Finally, I have a feeling that there
is unused code which can be deleted as it is a relic from the copy of
the J(w) mapping code and is not needed. For the 1.3 line code I
would recommend that identical functions are shifted into files such
as 'specific_fns/base_class.py', but for the 1.2 line code I would
prefer the duplication as this means that the current stable code base
remains stable.
Cheers,
Edward
On 6/26/07, Sebastien Morin <sebastien.morin.1@xxxxxxxxx> wrote:
Hi,
I started working on implementing the consistency tests last week
before
the last post was made and, hence, I worked on repository line 1.2
(revision 3303).
I implemented the consistency tests as a new type of run ('ct')
similar
to the one for Jw mapping.
The calculations are made for J(0), F_eta and F_R2 separately for each
magnetic field (one at a time). The output results file is similar to
the one for Jw mapping. The user then needs to plot them and look for
consistency with its own criteria (calculation correlation
coefficients, mean ratios and standard deviations, etc).
Please look at the followinr URL for a listing of the modifications to
old files and also necessary new files.
http://maple.rsvs.ulaval.ca/mediawiki/index.php/Relax_development
The file 'sample_scripts/consistency_tests.py' should be useful to
understand how the new procedure works.
Even if this was done on repository line 1.2, I think it is quite fine
since nothing was deleted but only things added (maybe too much, as I
reproduced the Jw mapping approach, maybe too much as I added lines in
the codes for grace, molmol, etc, maybe too much also since some
code is
duplicated from the Jw mapping code). The test-suite still works
perfectly and, so, I think it could be fine to add the tests to the
1.2
line as well... However, if necessary, I could implement the
consistency
testing procedure on line 1.3, following your comments as I am
quite new
to Python and maybe made things somehow not perfectly...
Please tell me what you think about this.
Cheers,
Sébastien :)
Edward d'Auvergne wrote:
Hi,
I have previously talked about data set consistency. For example
see
the post at
https://mail.gna.org/public/relax-users/2007-06/msg00001.html
in which a few reasons for inconsistencies have been explained. I
have, from experience, noticed that small changes in protein
concentration can change the collected relaxation rates
significantly
- most likely because of packing interactions. All samples should
essentially be identical in all respects for the relaxation rates to
be compared. And the temperate should always be fine tuned between
experiments and spectrometers using methanol (and always checked
later
on if there is a large time between collecting the same experiment).
Therefore these tests would be quite useful. Data consistency is
essential for the model-free results to be correct (as well as
reduced
spectral density mapping, SRLS, etc.) as this affects both the
optimisation and model selection and can result in artificial
motions
appearing. However I don't know how these test would currently fit
within relax. Maybe a new type of analysis should be created for
this
(see the pipe.create() user function in the 1.3 line or the
run.create() user function in the 1.2 line). These ideas should all
go into the 1.3 line (via a branch) as the 1.2 line is stable and no
new major features will be added to this code. What are the
ideas you
have been playing with?
Cheers,
Edward
On 6/15/07, Sebastien Morin <sebastien.morin.1@xxxxxxxxx> wrote:
Hi everyone
During the last months, I was astonished to realize that some spin
relaxation data I had acquired at different fields were not
consistent
between each other. The way I realized that was by seeing
discrepancy
between J(0) values calculated with those different datasets.
I looked a little bit in the litterature and found some interesting
consistency tests in a paper by Fushman (Fushman et al., JACS,
1998,
120:10947-10952).
This paper present 2 consistency tests to compare datasets from
different magnetic fields / samples / time / etc.
I think it would be interesting to implement those simple tests
in relax
so the user can, before trying to fit their data, know the
quality of
those... Regrettably, very few people look at the consistency of
their
datasets before analysis...
The underlying principle is the same as when looking at
consistency for
J(0). Thus, I think that those two tests and a J(0) test should be
implemented altogether...
I'll try to work a bit on this. Mimicking the code for spectral
density
should be a good starting point.
Am I right ?
Do you see any value in those tests ?
Cheers
Sébastien :)
_______________________________________________
relax (http://nmr-relax.com)
This is the relax-devel mailing list
relax-devel@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-devel
--
______________________________________
_______________________________________________
| |
|| Sebastien Morin ||
||| Etudiant au PhD en biochimie |||
|||| Laboratoire de resonance magnetique nucleaire ||||
||||| Dr Stephane Gagne |||||
|||| CREFSIP (Universite Laval, Quebec, CANADA) ||||
||| 1-418-656-2131 #4530 |||
|| ||
|_______________________________________________|
______________________________________
Re: Consistency tests