# Binomial errors

Hi,

I’m having trouble understanding how ROOT calculates binomial errors. I divide two histograms using the following code:

TH1D *run1_pfrac_Inner= new TH1D(“run1_pfrac_Inner”,“Run1”, 20, 20, 120);
run1_pfrac_Inner->Sumw2();
run1_pfrac_Inner->Divide(run1_pInner_allClu, run1_pInnerPitch, 1, 1, “B”); run1_pfrac_Inner->Draw();

I have a point with a probability of one. ROOT tells me the error on this point is 0.08. Have is this error calculated?

Cheers,

Lisa

In the case of points with p = 1, the binomial error is calculated using the Baysian approach, like in TGraphAsymmErrors::BayesDivide

root.cern.ch/root/htmldoc/TGraph … ayesDivide

I personally think, this is not very consistent, because for p= 1-epsilon the error is close to zero and then a non zero value value for p = 1.

Better using TGraphAsymmErrors:BayesDivide if you want Bayes errors, or alternatively use TH1::Divide for standard binomial errors (error - 0 for p = 0)

Best Regards,

Lorenzo