I have been fitting data using a function defined with 10 measured parameters and 2 free ones that I want to read out.
So far I have assigned one of the measured parameters, say mp to the bin at -1 which is not otherwise used, giving its measured value and measured error to that bin and making the fit function give the value of mp it is using to that bin, and I left all the other measured parameters fixed so far.
The bin thing works and I could maybe just keep doing that but it would be messy to clean up plot outputs, I don’t know if it will work properly if I use a log likelihood fit, and it does not allow me to incorporate the co-variance of the 10 parameters in to the chi-squared. Is there a more reliable way to modify the chi-squared used for fitting depending on deviations from measured values of parameters? I was not able to find one so far.
The lines relevant to the fit are
TF1* f1 = new TF1(“CaliT”,mCalTime,-1,20,4);
TFitResultPtr fitres = multH->Fit(“CaliT”,“S”,"",-1,17)
as a separate function:
Double_t mCalTime(Double_t* x, Double_t* par)
This is the function where I currently pick values of a parameter fit them in the bin at -1.