I’m trying to use TMultiDimFit to provide an interface to fit functions of ~arbitrary dimension specifically with the example of an EFT fit in mind.
TMultiDimFit seems to do exactly this for me in a very neat and tidy way, but there is one feature I haven’t been able to find. If I’ve understood correctly, the fit is not run over the raw data but on the data scaled to [-1,1], therefore the fitted coefficients correspond to that scaled data. But ultimately what I want the coefficients for the raw data. I suppose these can be extracted with some inversion, but would it be possible to get TMultiDimFit to do this for the user?
The class (ROOT: TMultiDimFit Class Reference) contains functions like MakeCode or MakeMethod allowing to write the code for evaluating the method in a general way. Have you tried those functions ?
Yes, I’ve seen these functions, but I was hoping to dynamically use the fit result within a much larger framework. To do that with a static .C/.cxx file would not be possible. And in any case the outputted code only performs the function evaluation, not any conversion of coefficients. So unless I’m missing something it’s they’re not going to do what I need… perhaps I can rework for my needs, but I was just thinking it would be natural for TMultiDimFit to do this for the user…
FYI, I think I’ve found a way to do what I want with ROOT::Fit. I still think adding this could be useful, but from my side I wouldn’t say it was essential…