IntegralAndError vs GetBinError (loop over bins)

ROOT
1

Hi all,

Yesterday I found something strange, it maybe because I made some mistakes in my code but I couldn’t figure it out (at least, until now).

I filled my histograms with sumw2 (SetDefaultSumw2(1)), then I calculate the error by two methods:

1/ IntegralAndError

2/ Calculate the sum of GetBinError of all bins.

However, I got different results.

Here is my code:

Int_t nbins = totalBG->GetSize(); Double_t bgerror; Float_t total = totalBG->IntegralAndError(1,nbins-2,bgerror,""); Double_t bgerror1 = 0; for (Int_t i=1;i<nbins-1;i++) { bgerror1 += histos[0]->GetBinError(i);}

Any help would be welcome, thank you guys so much.

2

In the “for” loop:
bgerror1 += (histos[0]->GetBinError(i)) * (histos[0]->GetBinError(i));
Afterwards:
bgerror1 = TMath::Sqrt(bgerror1);
See http://root.cern.ch/root/html/src/TH1.cxx.html#UdxrFB and http://root.cern.ch/root/html/src/TH1.cxx.html#xQJNQB

3

thank you for your quick reply, I’ll try it

4

Suppose I want to calculate something like the error of number of signal over number of signal, i.e calculate the error of the integral over the integral of the histogram, am I doing it right?

I’m trying to calculate the error of the significance here.

float significance = sqrt(2 * ((Signal + Background) * log(1 + (Signal/Background)) - Signal)); float error = significance * sqrt(pow((errorSignal/Signal),2) + pow((errorBackground/Background),2));