[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Help-gsl] non-linear least squares fitting

From: Harroun, Thad - NRC
Subject: [Help-gsl] non-linear least squares fitting
Date: Thu, 7 Apr 2005 16:45:35 -0400

In the non-linear least squares fitting, is there no way to hold a parameter
fixed? My strategy has been to declare a global array of parameters and
flags. If in the gsl_multifit_function_fdf evaluation, the fix flag is found
to be set, then use the global (original) parameter value, instead of the
one supplied by the fdfsolver.

The problem with this is the fdfsolver continues to change the parameter,
but can't know that the returned values (f and J) were evaluated using the
fixed parameter. Does this send the fdfsolver on a wild chase, and is the
chi-squared or covariant matrix wrong in the end? 

Has anyone else found a way around this problem?

PS. The manual text on the non-linear least squares fitting follows from the
section on multidimensional minimization.  But in this case, I find the
notion of parameters and x values to be backwards. Most would expect "x" to
be the name of the data point, and not the parameter to be minimized.  

Canadian Neutron Beam Centre
National Research Council Canada
Building 459, Station 18
Chalk River Laboratories
Chalk River, Ontario
Pho:613-584-8811 x6237

reply via email to

[Prev in Thread] Current Thread [Next in Thread]