Which TRolke method is correct for determining the upper limits of a signal for the following case:
- Nsig and Nbkg are determined from the fit (RooFit)
- signal PDF shape is fixed from signalMC
- signal is very small
- and signal efficiency is determined independently from signalMC
I am aware of different class methods: SetPoissonBkgBinomEff, SetPoissonBkgGaussEff, etc… But which one would be the right one in this case?
ROOT Version: 6.24/06
Platform: Jupyter Notebook
Dear @AndrejL ,
Thank you for reaching out to the forum! I believe @moneta can help you out here.
You can use
TRolke only for a simple Poisson counting process, where you observe n events and you have Nsig expected events and Nbkg from background.
If you have values obtained from a fit using RooFit, you should use RooStats to compute confidence interval.
TRolke uses the profile likelihood method, so you can use in this case the ProfileLikelihood calculator from RooStats (see this tutorial).
Instead if you need to compute an upper limit is better to use the Asymptotic calculator from RooStats.
(see this other tutorial).