Frida contains a quite extensive set of routines to fit data.
For fitting of the data a Gauss-Newton algorithm is employed at the moment.
Fitting of the data makes use of the E04FCF
routine of the
NAG FORTRAN library.
Frida allows to adjust a number of parameters such as the tolerance of the fit
etc.. The following page lists a description of the different commands available
in the curve fitting menu c.
ca specifies auxiliary parameters, which are the fit range, and wether
the fit shall be weighted with stepwidth in x, error bars associated to the
data points, or logarithmically. The latter takes especially into account if
data is varying over several orders of magnitude as it is typically the case for
scattering laws measured by means of quasielastic neutron scattering. Moreover, a
flag can be set to specify that the analytical curve shall be convolved with a
another curve (e. g. resolution function).
ci allows to extract the fit parameters from the fit curve saving them into
an integral file. Imagine the following: The scattering law S(q,w) has been
measured for several different q-values. S(q,w) is fit to a model function
e. g. a Lorentzian. The linewidth HWHM is one of the fit parameters that varies
with q. ci extracts this parameter with q as the x-grid and HWHM as
cg creates a file that contains the curve on a grid which can be specified to be
the same as the data-file, linear, logarithmical, semi-lograthmic or user-specified.
This is e. g. useful if one wants to subtract the fit-curve from the data file or the direct display of the function yields oscillations.