Wrong TGraphAsymmErrors error length

[jerryliu]

Hello,

I plot the error bar of the efficiency but I think the length of error is not correct.
My test code is here and I have to use BayesDivide to derive efficiency.
The error bar seems too large and I am not familiar with Bares method.
Is the result correct?

#include "TGraphAsymmErrors.h"
#include "TCanvas.h"
#include "TH1F.h"
#include "TMath.h"

using namespace std;
void aserror() {
    TCanvas *c1 = new TCanvas("c1","c1",3);
    TH1F* h_pass = new TH1F("h_pass","h_pass",5,0,5);
    TH1F* h_all = new TH1F("h_all","h_all",5,0,5);
    h_pass->Sumw2();
    h_all->Sumw2();
    for (int i=0;i<5;i++) {
        h_pass->Fill(i,TMath::Power(10,i+1)/2);
        h_all->Fill(i,TMath::Power(10,i+1));
    }   
    TGraphAsymmErrors* tgas = new TGraphAsymmErrors(5);
    tgas->BayesDivide(h_pass,h_all);
    tgas->Draw("ALP");
}

c1.pdf (13.3 KB)

---

_ROOT Version: ROOT 6.12/06
_Platform: Ubuntu 14 LTS
Compiler: Not Provided

---