Number of independent points: Artemis & Feffit
Hi guys,
I've noticed a strange behaviour in Artemis during the calculation of
N_idp; the value differs from the Feffit one. In fact, with command line
feffit (ver. 2.984) N_idp=(2*delta_k*delta_R/pi)+2 but with Artemis the
"+2" value disappears. Seen that the procedure is:
Artemis-->Ifeffit-->feffit(), why they differs?
Mauro
--
Mauro Rovezzi
On Tuesday 28 March 2006 04:50, Mauro Rovezzi wrote:
Hi guys,
I've noticed a strange behaviour in Artemis during the calculation of N_idp; the value differs from the Feffit one. In fact, with command line feffit (ver. 2.984) N_idp=(2*delta_k*delta_R/pi)+2 but with Artemis the "+2" value disappears. Seen that the procedure is: Artemis-->Ifeffit-->feffit(), why they differs?
Mauro, Artemis needs to be able to report on Nidp at any time, even before any commands have been sent to Ifeffit. Consequently, it computes Nidp by itself based on the current state of the project. You are correct that it reports N_idp=(2*delta_k*delta_R/pi) -- I just looked at the code and that is indeed what it does. I don't think, however, that you are correct that Ifeffit adds the 2. I also looked at the Ifeffit code. At line 341 in iff_feffit.f, Nidp is computed without the added constant. (But perhaps I am missing something else in the code...?) I just ran a fit with Artemis -- Ifeffit's value of n_idp in that fit was 18.56 and Artemis reported 19. It is true that there is a difference in how Ifeffit computes Nidp during a fit and how Artemis computes it when asked. Ifeffit uses that actual grid values of the first and last points of the actual data grid in the Fourier transform range in k and of the fit range in R. Artemis uses the numbers entered into the k-range and R-range boxes on the data page. It is not unbelievable that those values might conspire to be different by 1. B -- Bruce Ravel ---------------------------------------------- bravel@anl.gov Molecular Environmental Science Group, Building 203, Room E-165 MRCAT, Sector 10, Advanced Photon Source, Building 433, Room B007 Argonne National Laboratory phone and voice mail: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://cars9.uchicago.edu/~ravel EXAFS software: http://cars9.uchicago.edu/~ravel/software/exafs/
Bruce Ravel wrote:
You are correct that it reports N_idp=(2*delta_k*delta_R/pi) -- I just looked at the code and that is indeed what it does. I don't think, however, that you are correct that Ifeffit adds the 2. I also looked at the Ifeffit code. At line 341 in iff_feffit.f, Nidp is computed without the added constant. (But perhaps I am missing something else in the code...?) It is true that there is a difference in how Ifeffit computes Nidp during a fit and how Artemis computes it when asked. Ifeffit uses that actual grid values of the first and last points of the actual data grid in the Fourier transform range in k and of the fit range in R. Artemis uses the numbers entered into the k-range and R-range boxes on the data page. It is not unbelievable that those values might conspire to be different by 1.
In my case the difference (+2 value) is with the command line program
"feffit". I mean, giving a feffit.inp file to feffit I obtain a
different result than using Ifeffit/Artemis with the same paths/variables.
To be more precise, my question should be: why differs the minimization
feffit() function in Ifeffit with the "feffit" command line program?
M.
--
Mauro Rovezzi
Hi Mauro,
In my case the difference (+2 value) is with the command line program "feffit". I mean, giving a feffit.inp file to feffit I obtain a different result than using Ifeffit/Artemis with the same paths/variables.
To be more precise, my question should be: why differs the minimization feffit() function in Ifeffit with the "feffit" command line program?
Yes, the calculation of the number of independent points changed between Feffit and Ifeffit. It was decided to be more conservative and to avoid the potential abuses of the '+2' issue. With the older calculation, it is "allowed" to fit 2 parameters with Delta k = 0.1 and Delta R = 0.1. That might give the impression that one can use EXAFS tp determine distance and coordination number without any data. EXAFS analysis often has a bad reputation, so I hope you'll forgive me for not wanting to give that impression. For backward compatibility, and the benefit of the few Feffit users (most of whom are assumed to know about these issues and act responsibly to avoid them) Feffit was not changed.. --Matt
participants (3)
-
Bruce Ravel -
Matt Newville -
Mauro Rovezzi