P.d.f value is Not-a-Number in self-made Gaussian pdf

[chanchal]

Dear Expert,

I want to do a 2D fit using a self-made Gaussian pdf (myRooGaussian1stbin) which is normalized in the fitting range. But with free any of the parameters, this pdf value is pushed to nan (below).

p.d.f value is Not-a-Number (-nan), forcing value to zero @ x=pi0_m_prefit=0.1788, a_mu=a_mu=-0.873702, b_mu=b_mu=1.02511, c_mu=c_mu=3.36042e+08, d_mu=d_mu=4197.5, p0_sigma=p0_sigma=-nan, p1_sigma=p1_sigma=-nan, y=pi0_errm=0.00993388

Here is the code that I am using:
2D_fit_withselfpdf.zip (79.8 KB)

Can you please suggest where I am wrong?

Regards
Chanchal

[moneta]

Hi

You should check that both the pdf un-normalized value is not zero and also that its integral is not zero. If they are this will cause a NaN when evaluating the pdf

Lorenzo

[chanchal]

Hi @moneta,

Yes, I have checked this with a free parameter of sigma but pdf un-normalized value (value) and its integral (norm) both are also pushed to nan.
I don’t understand why this happens.

Code are attached here:
2D_fit_selfpdf_new.zip (79.8 KB)

Can you please help me with that?

Regards
Chanchal

[moneta]

Hi,
The problem is exactly what I have said before, you are having zero integral and zero pdf.
Running your code you can see this message:

RooAbsPdf::gauss[ x=pi0_m_prefit a_mu=a_mu b_mu=b_mu c_mu=c_mu d_mu=d_mu p0_sigma=p0_sigma p1_sigma=p1_sigma y=pi0_errm ]
     p.d.f normalization integral is zero or negative: 0.000000 @ x=pi0_m_prefit=0.1796, a_mu=a_mu=-0.873702, b_mu=b_mu=1.02511, c_mu=c_mu=3.36042e+08, d_mu=d_mu=4197.5, p0_sigma=p0_sigma=1.34232 +/- 0.0732578, p1_sigma=p1_sigma=-56.3534, y=pi0_errm=0.0052062

Probably the parameter values above do not make sense. You should maybe add bounds on some of the parameter values to exclude them.