Some questions about Likelihood

Roofit and RooStats
1

Hi all

I’am doing a fit using roofit and the maximum likelihood. The fit result gives some parameters that I don’t understand.
The first is the EDM( estimated distance to minimum). My fit gives 2.64592e-05. I don’t know if it’s good or not.The second is the - Log (Likelihood), the value returned is -90999.

Other point: when you use the likelihood, the chi2/ndf is a good proof of the goodness of the fit ?

Thanks

2

Hi,

lower value of EDM means better convergence of the fit. However, you need to have small numerical error in your log-likelihood function evaluation in order to reach very small values. Normally a value of EDM<1.E-6 is required for a fit to converge.

Chi2/ndf of your binned data vs the function could be used as quality of the fit, while the likelihood value can’t be used.
Lorenzo