Issue with Proper Error Handling in RooFit Binned Data Fitting

Roofit and RooStats
1

Dear all,

When performing binned data fitting in RooFit, the errors are typically treated as Poissonian by default. However, the actual errors in my binned data are significantly larger than the Poisson expectation.

To investigate this, I conducted a test using the following script, but both chi2fitTo and fitTo() yield identical fitting errors, which is clearly unphysical. Enabling SumW2Error(true) also fails to resolve the issue.

How can I obtain statistically correct fitting errors that properly account for these larger uncertainties?
chi2fit.cxx (2.7 KB)

2

Hi,

Thanks for posting.
I add the RooFit expert in the loop, @jonas , but mind, the answer might come with some latency.

Best,
D

3

Thank you! Your expertise on Roofit would be greatly appreciated—no rush, and we appreciate your time whenever you’re able to weigh in.