[RooFit] Goodness-of-fit in 2-d binned likelihood fit

ROOT
1

Hi, all

I fit 2-dimensional binned likelihood fit to the data with uncorrelated 2 variables.
To show goodness-of-fit after fitting, I plot the data and fit function on each dimension.
Then I got -2log(LR),LR means Likelihood Ratio, using data points and function curve in each dimension and summed them. I took the number of bin in two dimensions minus the number of floated parameter from minuit as the number of degree of freedom(ndf). Finally I showed -2log(LR)/ndf as goodness-of-fit.

Is it reasonable?

Can we call 2-dimensional likelihood fit as simultaneous fit for 2-dimension if each dimension is not correlated?

Thanks in advance

2

Hi,

Sorry for the late reply, I was offline last week.

Your question is more of a statistics rather question than a RooFit question, to which I have unfortunately no good answer. You should consult a (local) statistics expert as multi-dimensional goodness-of-fit really is non-trivial statistical issue to which there are often no easy solutions.
Wouter