I have events that I correct for some effects by giving them a weight. Then I call for each event Fill(value, weight) (after Sunw2()).
My issue is that I didn’t find a ROOT way to make the error computation taking into account error on the weights themselves. Is there a way to do so? Something like Fill(value, weight, sigma_weight) ?
When computing uncertainties with weights it is assumed they do not have uncertainties and only the statistical flactuations of observing an event with a given weight is computed.
If you have significant weight uncertainty you need to model correctly your problem including a-priori uncertainty in the weights. This uncertainty might also depend on the observed value.
For this case you need to estimate yourself the global uncertainty