Unbinned fitting (Likelihood)

[bdrum]

For fitting in #44 I have standard $\chi^2$ test for deterring quality of the fit, but for area with small statistics (Puasson) it's wrong.
E.g. I've calculated error as $\sqrt(N)$ for each bin, but in case of small count of events in bin it provide us the situation when 2 events in bin and 1.4 sigma.

Here is the fragment from lection of Cambridge prof. Mark Thomson about the question: