TH1F bin content saturation

[stwunsch]

See for comparison the screenshot below.

The upper plot was done with TTree.Draw:

>>> import ROOT
>>> f = ROOT.TFile('DYJetsToLL.root')
>>> t = f.Get('Events')
>>> t.Draw('GenPart_pdgId')

The lower plot was done with RDataFrame.Histo1D:

>>> import ROOT
>>> c = ROOT.TCanvas()
>>> h = ROOT.RDataFrame('Events', 'DYJetsToLL.root').Histo1D('GenPart_pdgId')
>>> h.Draw()

I've used ROOT 6.22/02 and you can download the file here:

http://opendata.web.cern.ch/record/12353

[bellenot]

@stwunsch which one is correct?

[stwunsch]

Good question! I guess the lower one, also checked by counting only the respective entries in the GenPart_pdgId vector.

[stwunsch]

Well, it's a DY sample, so physically spoken we would expect the upper one (?) Let's say, it's not so clear what's going on 😅

[eguiraud]

Can you enforce the same exact binning in both and also try with raw TTree and raw TTreeReader? If RDF or TTree is producing wrong histograms it's a critical issue if not a release blocker.

[stwunsch]

@eguiraud Do we agree that the code above should produce (in principle) the same output?

[stwunsch]

@eguiraud Here you can find the plots produced with the same binning.

>>> import ROOT
>>> f = ROOT.TFile('DYJetsToLL.root')
>>> t = f.Get('Events')
>>> h = ROOT.TH1F('h', 'h', 100, -20, 20)
>>> c = ROOT.TCanvas()
>>> t.Draw('GenPart_pdgId>>h')

>>> import ROOT
>>> c = ROOT.TCanvas()
>>> h = ROOT.RDataFrame('Events', 'DYJetsToLL.root').Histo1D(('h', 'h', 100, -20, 20), 'GenPart_pdgId')
>>> h.Draw()

[stwunsch]

As a cross check: The manual count in RDF of the entries we see in the histogram.

>>> ROOT.RDataFrame('Events', 'DYJetsToLL.root').Define('n', 'Sum(GenPart_pdgId == 11)').Sum('n').GetValue()
36148151.0
>>> ROOT.RDataFrame('Events', 'DYJetsToLL.root').Define('n', 'Sum(GenPart_pdgId == 13)').Sum('n').GetValue()
25642431.0
>>> ROOT.RDataFrame('Events', 'DYJetsToLL.root').Define('n', 'Sum(GenPart_pdgId == 15)').Sum('n').GetValue()
18289536.0

[eguiraud]

Do we agree that the code above should produce (in principle) the same output?

I don't see why not. We really need an authoritative answer as to which of the plots is correct though. Maybe raw TTree and TTreeReader can act as tie-breaker. There are no missing entries (the number of entries in the two histograms is the same) so it seems that one of the two interfaces is reading wrong values -- but how can it be that histogram mean and variance are exactly the same in the two cases? The mean, I can understand, but the variance should be larger for the upper histo than for the lower histo.

[stwunsch]

Here a reproducer with my best raw TTree skills:

void test() {
    auto f = TFile::Open("DYJetsToLL.root");
    auto t = f->Get<TTree>("Events");
    UInt_t n;
    Int_t x[100];
    t->SetBranchAddress("nGenPart", &n);
    t->SetBranchAddress("GenPart_pdgId", x);
    auto n11 = 0u;
    auto n13 = 0u;
    auto n15 = 0u;
    for(auto i = 0u; i < t->GetEntries(); i++) {
        t->GetEntry(i);
        for (auto j = 0u; j < n; j++) {
            if (x[j] == 11) n11++;
            if (x[j] == 13) n13++;
            if (x[j] == 15) n15++;
        }
    }
    cout << "n11: " << n11 << endl;
    cout << "n13: " << n13 << endl;
    cout << "n15: " << n15 << endl;
}

n11: 36148151
n13: 25642431
n15: 18289536

So we can conclude that's a plotting issue in TTree::Draw? And I think RDataFrame does the right thing.