I am trying to fit a sum of two histograms where the coefficient for one of them is negative. I am using RooRealSumPdf to add them. I declare the coefficient for one of them and expect that, according to docs, the other coefficient will be calculated. However, when I supply only one coefficient, the fitter doesn’t work. To make it work I need to supply both coefficients.
Below’s an MWE:
x=RooRealVar('x','x',0,1)
f=RooGenericPdf('f','f','x-x^2',x)
y=f.generate(x,10000)
f1=RooFormulaVar('f1','f1','x',x)
f2=RooFormulaVar('f2','f2','x^2',x)
coef1=RooRealVar('coef1','coef1',1,0.9,1.1)
coef2=RooRealVar('coef2','coef2',-1,-1.1,-0.9)
The fitter works for the line below:
f_fit=RooRealSumPdf('f_fit','f_fit',(f2,f1),(coef2,coef1))
I get the correct values of coefficients:
RooRealVar::coef1 = 1.00018 +/- 0.0239468 L(0.9 - 1.1)
RooRealVar::coef2 = -0.999821 +/- 0.0239383 L(-1.1 - -0.9)
But the fitter doesn’t work if I add the functions with just one coefficient
f_fit=RooRealSumPdf('f_fit','f_fit',(f2,f1),(coef2,))
I get the error RooAbsMinimizerFcn: Minimized function has error status. and the value for coef2 tends toward a value close to zero. If I let the range of coef2 be positive, it always gives a positive number close to zero.
ROOT version 6.26/14