I am creating an efficiency plot by dividing two histograms of weighted events (all events have different weights). In the division of histograms I put the “B” option to get the binomial error.
I would like to know how this binomial error is calculated when the events have a weight?
In TH1::Divide with option “B” the error is computed using the “wrong” gaussian approximation and then using error propagation. This is not fully correct. I hope to introduce a better error calculation in the case of weights in the TEfficiency class. Currently the class does not support weighted events
Just to let you know that I have implemented in the ROOT trunk the possibility to use histogram with weights in TGraphAsymmErrors::Divide(h1,h2);
Only the Bayesian and the normal approximation methods are so far supported. The normal approximation is the same used in TH1::Divide
The TEfficiency class will be improved later since it requires some changes in the interface