Goodness of fit

I have two histograms of the Co60 beta decay spectrum with me. One of them is simulated using geant4 and the other one is experimentally obtained. I have scaled the GEANT4 spectrum to the experimental count rate and visually they seem a good fit. But I would like to perform a quantitative measurement of the goodness of the fit. I could have performed the chi-square test or reduced chi-square test in root, but the problem is that both the histograms are having different binning. As far as I know, the chi-square fit works if the binnings are the same. Is there any other option in root for estimating the goodness of fit with different binning?

The data is stored as the bin center- bin content format in data files as follows
13.txt (7.1 KB)
14.txt (10.6 KB)

_ROOT Version:_ROOT 6.26/10
Platform: ubuntu~22.04
Compiler: gcc

Hi Gokul,

You have actually 2 issues. As you pointed out your binning is different but also both histograms have (counting) uncertainties.

The different binning could be addressed by interpolating your high-statistics (simulation) data points to the centers of the experimental data points. The “new” simulation data point is a function of the neighboring ones and in case the points are close, a linear interpolation should suffice. The “new” error follows from straightforward error propagation.

Issue two could be tackled with the TFractionFitter class.

Limit this analysis to x < 1 because above it both histograms get sparse.


1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.