When fitting mathematical models to data, it is often not enough to simply fit one dataset; it is necessary to globally fit many sets simultaneously, for which some of the model parameters are shared, and others are per set.
Does Numerics have capability for solving this problem? If not (at least I don’t see it), is this planned? Or maybe someone knows an example of a working framework (classes, methods …) to implement such a global fit.
I am currently using the following algorithm. Data from different datasets are combined into one contiguous dataset with a shift in the independent variable so that each subsequent dataset starts where the previous one ends. To fit such a combined set, a fitting function is used that takes parameters combined of a set of the shared parameters (common for the entire data group), and sets of individual parameters for each section, plus fixed parameters corresponding to the shifts of individual sections in this common set. So far, I have not brought this algorithm to a universal form, but I am implementing it as needed, in special cases. Maybe this could serve as an idea for someone who can find the time to develop the framework.
1 post - 1 participant