Is it possible to use interpolating function with parameters as a model for
NonLinearModelFit (and other integrated methods for fitting)?
Similar question was already posted earlier - How do I find the best parameter to fit my data if the model is a interpolating function?. However, answer is not covering whole fitting procedure and it is based on "do it yourself" method, since the question was only about the value of best fit parameter.
Difference is that I am asking how to use (if possible) Mathemaca's methods for fitting.
NonLinearModelFit finds not only best values of parameters, but also $\chi^2$, confidence regions, correlation matrix and many more things that I use. I could calculate them on my own, but I want to use some of integrated methods.
If one defines a model in the form
model[x, A, B, C], which gives a real number for a set of parameter values
A, B, C and
a value of
x, then Mathematica knows how to play with it. Instead I have a model defined as
model[A, B, C], which gives, for any set of numeric values A, B and C, a pure function or a function of
x on some range. Now how to explain that to Mathematica and to use
NonLinearModelFit? One thing I could do is to redefine pure function
model[A, B, C] to
model[x, A, B, C], but that would require calculating the whole slope
model[A, B, C] while taking its value at only one point
x, and then repeating this procedure for every point
x. This takes too long if one has big set of data points (with errors) to fit.
Is there a better idea?