Hi All,
I’m trying to do a weighted unbinned ML fit using RooFit. The fit function I am using is Double Sided Crustal Ball + Gaussian(same mean). The weights per event range between [2.6950914e-11, 3.7362415e-08] . As I can see the initial values and range of the parameters remarkably affect the fit results. In the figure attached , fig2 is made with just changing the fraction variable’s(f4) initial value (code Modeling.py attached). As can be seen the fitted value for f4 is not much different between both the plots but the other variable values and uncertainities change ( eg. the value for ZH_sig_s). I have also tried slightly changing the initial values and range for other variables too and everytime i get different results. Thus I am not sure whether i am doing the fit correctly. Following are my queries:
- How to choose the initial values and range of the parameters so that the fit result is stable
- For this case is it getting unstable because of very low weights of the events? In that case how should i make up for that. (AsymptoticError is enabled during the fit. I have tried using SumW2Error also)
- The chi2 shown in the plots is the value: xframe.chiSquare(8), 8 being the no of floating parametrs.
Is this method to quantify chi2 correct? The value wiill change when varying the no of bins in the frame.
Any comment or suggestion is highly welcomed.
Thanks in advance.
Modeling.py (2.5 KB)