Trouble with Integral() method for ROC curves

Newbie
1

I am trying to calculate the integral of my ROC curves for background rejection vs signal efficiency using the Integral() method described in the ROOT TH1:Integral reference guide.
Instead of returning values between 0 and 1. For values around 0.5, indicating a straight line ROC curve, I am instead receiving values around 0 with values trending upwards towards some value from there. I am not sure why the integral method would read out 0 for what is basically a triangle.

Here is the code I am using:

                    // Calculate Background Vs. Signal Efficiency
                    int nSigBins = hist->GetNbinsX();
                    int nBkgBins = bkghist->GetNbinsX();

                    double TSigIntegral = hist->Integral();
                    double TBkgIntegral = bkghist->Integral();

                    std::vector<double> LeftSigEff(nSigBins, 0.0);
                    std::vector<double> RightSigEff(nSigBins, 0.0);
                    std::vector<double> LeftBkgEff(nBkgBins, 0.0);
                    std::vector<double> RightBkgEff(nBkgBins, 0.0);

                    for (int i = 0; i < nSigBins; i++) {
                        LeftSigEff[i] = (hist->Integral(0, i) / TSigIntegral);
                        RightSigEff[i] = (hist->Integral(i, nSigBins) / TSigIntegral);
                    }

                    for (int i = 0; i < nBkgBins; i++) {
                        LeftBkgEff[i] = 1 - (bkghist->Integral(0, i) / TBkgIntegral);
                        RightBkgEff[i] = 1 - (bkghist->Integral(i, nBkgBins) / TBkgIntegral);
                    }

                    // Plot Efficiency Comparison of Background and Signal
                    TCanvas *LeftCanvas = new TCanvas("LeftCanvas", "LeftCanvas", 800, 600);
                    TLegend *LeftLegend = new TLegend(0.6, 0.6, 0.9, 0.9);
                    LeftCanvas->cd();
                    
                    auto g = new TGraph(nSigBins, &LeftBkgEff[0], &LeftSigEff[0]);
                    LeftCanvas->SetTitle("Left Cut Efficiency Comparison");
                    g->GetXaxis()->SetTitle("#epsilon_{Sig}");  
                    g->GetYaxis()->SetTitle("1 - #epsilon_{Bkg}");
                    g->GetYaxis()->SetRangeUser(0, 1);
                    g->GetXaxis()->SetRangeUser(0, 1);
                    std::cout << "----------------------------------------" << std::endl;
                    std::cout << "Left Cut Efficiency Comparison" << std::endl;
                    std::cout << "Decay Channel: " << decayinfo << std::endl;
                    std::cout << "Observable: " << obs << std::endl;
                    std::cout << "Particle: " << part << std::endl;
                    std::cout << "ROC Curve Plot Integral: " << 100*g->Integral() << std::endl;
                    g->Draw("AC*");
                    LeftCanvas->SaveAs(("Output/RequestAnalysis/AllSMass/" + decayinfo + "_" + part + "_" + obs + "_LeftEfficiencyComparison.png").c_str());

Where hist and bkghist are histograms loaded in separately.

Here is some of my output:
Left Cut Efficiency Comparison
Decay Channel: ttbb
Observable: jet
Particle: N
ROC Curve Plot Integral: 31.4656

Left Cut Efficiency Comparison
Decay Channel: ttbb
Observable: jet
Particle: pt
ROC Curve Plot Integral: 2.94997

Left Cut Efficiency Comparison
Decay Channel: ttbb
Observable: E
Particle: met
ROC Curve Plot Integral: 27.6885

Left Cut Efficiency Comparison
Decay Channel: ttbb
Observable: E
Particle: st
ROC Curve Plot Integral: 2.8683

Left Cut Efficiency Comparison
Decay Channel: ttbb
Observable: N
Particle: tau
ROC Curve Plot Integral: 0

I multiplied the output by 100 simply for readability.

_ROOT Version: 6.24/06
_Platform: lxplus
Compiler: Not Provided

2

The histogram bins go from 1 to nbins (included), not 0 to nbins-1 (bin = 0 and nbins+1 are under/overflow), so that may be an issue in your loops and integrals.