Show how to express ROOT's standard H1 analysis with RDataFrame.
Author: Axel Naumann, Danilo Piparo (CERN)
This notebook tutorial was automatically generated with ROOTBOOK-izer from the macro found in the ROOT repository on Thursday, March 23, 2023 at 10:45 AM.
%%cpp -d
auto Select = [](ROOT::RDataFrame &dataFrame) {
using namespace ROOT;
auto ret = dataFrame.Filter("TMath::Abs(md0_d - 1.8646) < 0.04")
.Filter("ptds_d > 2.5")
.Filter("TMath::Abs(etads_d) < 1.5")
.Filter([](int ik, int ipi, RVecI& nhitrp) { return nhitrp[ik - 1] * nhitrp[ipi - 1] > 1; },
{"ik", "ipi", "nhitrp"})
.Filter([](int ik, RVecF& rstart, RVecF& rend) { return rend[ik - 1] - rstart[ik - 1] > 22; },
{"ik", "rstart", "rend"})
.Filter([](int ipi, RVecF& rstart, RVecF& rend) { return rend[ipi - 1] - rstart[ipi - 1] > 22; },
{"ipi", "rstart", "rend"})
.Filter([](int ik, RVecF& nlhk) { return nlhk[ik - 1] > 0.1; }, {"ik", "nlhk"})
.Filter([](int ipi, RVecF& nlhpi) { return nlhpi[ipi - 1] > 0.1; }, {"ipi", "nlhpi"})
.Filter([](int ipis, RVecF& nlhpi) { return nlhpi[ipis - 1] > 0.1; }, {"ipis", "nlhpi"})
.Filter("njets >= 1");
return ret;
};
const Double_t dxbin = (0.17 - 0.13) / 40; // Bin-width
Definition of a helper function:
%%cpp -d
Double_t fdm5(Double_t *xx, Double_t *par)
{
Double_t x = xx[0];
if (x <= 0.13957)
return 0;
Double_t xp3 = (x - par[3]) * (x - par[3]);
Double_t res =
dxbin * (par[0] * pow(x - 0.13957, par[1]) + par[2] / 2.5066 / par[4] * exp(-xp3 / 2 / par[4] / par[4]));
return res;
}
Definition of a helper function:
%%cpp -d
Double_t fdm2(Double_t *xx, Double_t *par)
{
static const Double_t sigma = 0.0012;
Double_t x = xx[0];
if (x <= 0.13957)
return 0;
Double_t xp3 = (x - 0.1454) * (x - 0.1454);
Double_t res = dxbin * (par[0] * pow(x - 0.13957, 0.25) + par[1] / 2.5066 / sigma * exp(-xp3 / 2 / sigma / sigma));
return res;
}
Definition of a helper function:
%%cpp -d
void FitAndPlotHdmd(TH1 &hdmd)
{
// create the canvas for the h1analysis fit
gStyle->SetOptFit();
auto c1 = new TCanvas("c1", "h1analysis analysis", 10, 10, 800, 600);
hdmd.GetXaxis()->SetTitleOffset(1.4);
// fit histogram hdmd with function f5 using the loglikelihood option
auto f5 = new TF1("f5", fdm5, 0.139, 0.17, 5);
f5->SetParameters(1000000, .25, 2000, .1454, .001);
hdmd.Fit("f5", "lr");
hdmd.DrawClone();
}
Definition of a helper function:
%%cpp -d
void FitAndPlotH2(TH2 &h2)
{
// create the canvas for tau d0
auto c2 = new TCanvas("c2", "tauD0", 100, 100, 800, 600);
c2->SetGrid();
c2->SetBottomMargin(0.15);
// Project slices of 2-d histogram h2 along X , then fit each slice
// with function f2 and make a histogram for each fit parameter
// Note that the generated histograms are added to the list of objects
// in the current directory.
auto f2 = new TF1("f2", fdm2, 0.139, 0.17, 2);
f2->SetParameters(10000, 10);
h2.FitSlicesX(f2, 0, -1, 1, "qln");
// See TH2::FitSlicesX documentation
auto h2_1 = (TH1D *)gDirectory->Get("h2_1");
h2_1->GetXaxis()->SetTitle("#tau [ps]");
h2_1->SetMarkerStyle(21);
h2_1->DrawClone();
c2->Update();
auto line = new TLine(0, 0, 0, c2->GetUymax());
line->Draw();
}
TChain chain("h42");
chain.Add("root://eospublic.cern.ch//eos/root-eos/h1/dstarmb.root");
chain.Add("root://eospublic.cern.ch//eos/root-eos/h1/dstarp1a.root");
chain.Add("root://eospublic.cern.ch//eos/root-eos/h1/dstarp1b.root");
chain.Add("root://eospublic.cern.ch//eos/root-eos/h1/dstarp2.root");
ROOT::EnableImplicitMT(4);
ROOT::RDataFrame dataFrame(chain);
auto selected = Select(dataFrame);
Plugin No such file or directory loading sec.protocol libXrdSeckrb5-5.so
Note: The title syntax is "
auto hdmdARP = selected.Histo1D({"hdmd", "Dm_d;m_{K#pi#pi} - m_{K#pi}[GeV/c^{2}]", 40, 0.13, 0.17}, "dm_d");
auto selectedAddedBranch = selected.Define("h2_y", "rpd0_t / 0.029979f * 1.8646f / ptd0_d");
auto h2ARP = selectedAddedBranch.Histo2D({"h2", "ptD0 vs Dm_d", 30, 0.135, 0.165, 30, -3, 6}, "dm_d", "h2_y");
FitAndPlotHdmd(*hdmdARP);
FitAndPlotH2(*h2ARP);
FCN=10684 FROM MIGRAD STATUS=CONVERGED 209 CALLS 210 TOTAL EDM=1.2814e-06 STRATEGY= 1 ERROR MATRIX ACCURATE EXT PARAMETER STEP FIRST NO. NAME VALUE ERROR SIZE DERIVATIVE 1 p0 9.59915e+05 8.87690e+04 2.77938e+02 -4.39940e-08 2 p1 3.51114e-01 2.27896e-02 2.22437e-04 1.27888e-01 3 p2 1.18503e+03 5.92240e+01 3.18237e+00 1.06503e-05 4 p3 1.45569e-01 5.93973e-05 4.11621e-06 1.60533e+00 5 p4 1.24388e-03 6.60206e-05 3.72402e-06 -3.36599e+00 ERR DEF= 0.5
Draw all canvases
gROOT->GetListOfCanvases()->Draw()