Data and categories: tools for manipulation of (un)binned datasets
Author: Wouter Verkerke
This notebook tutorial was automatically generated with ROOTBOOK-izer from the macro found in the ROOT repository on Tuesday, March 19, 2024 at 07:16 PM.
%%cpp -d
#include "RooRealVar.h"
#include "RooDataSet.h"
#include "RooDataHist.h"
#include "RooGaussian.h"
#include "RooCategory.h"
#include "TCanvas.h"
#include "TAxis.h"
#include "RooPlot.h"
#include "TFile.h"
using namespace RooFit;
Binned (RooDataHist) and unbinned datasets (RooDataSet) share many properties and inherit from a common abstract base class (RooAbsData), that provides an interface for all operations that can be performed regardless of the data format
RooRealVar x("x", "x", -10, 10);
RooRealVar y("y", "y", 0, 40);
RooCategory c("c", "c");
c.defineType("Plus", +1);
c.defineType("Minus", -1);
RooDataSet is an unbinned dataset (a collection of points in N-dimensional space)
RooDataSet d("d", "d", RooArgSet(x, y, c));
Unlike RooAbsArgs (RooAbsPdf,RooFormulaVar,....) datasets are not attached to the variables they are constructed from. Instead they are attached to an internal clone of the supplied set of arguments
Fill d with dummy values
Int_t i;
for (i = 0; i < 1000; i++) {
x = i / 50 - 10;
y = sqrt(1.0 * i);
c.setLabel((i % 2) ? "Plus" : "Minus");
// We must explicitly refer to x,y,c here to pass the values because
// d is not linked to them (as explained above)
d.add(RooArgSet(x, y, c));
}
d.Print("v");
cout << endl;
DataStore d (d) Contains 1000 entries Observables: 1) x = 9 L(-10 - 10) "x" 2) y = 31.607 L(0 - 40) "y" 3) c = Plus(idx = 1) "c"
The get() function returns a pointer to the internal copy of the RooArgSet(x,y,c) supplied in the constructor
const RooArgSet *row = d.get();
row->Print("v");
cout << endl;
1) 0x7fdde9598230 RooRealVar:: x = 9 L(-10 - 10) "x" 2) 0x7fdde95f3560 RooRealVar:: y = 31.607 L(0 - 40) "y" 3) 0x7fdde91915f0 RooCategory:: c = Plus(idx = 1) "c"
Get with an argument loads a specific data point in row and returns a pointer to row argset. get() always returns the same pointer, unless an invalid row number is specified. In that case a null ptr is returned
d.get(900)->Print("v");
cout << endl;
1) 0x7fdde9598230 RooRealVar:: x = 8 L(-10 - 10) "x" 2) 0x7fdde95f3560 RooRealVar:: y = 30 L(0 - 40) "y" 3) 0x7fdde91915f0 RooCategory:: c = Minus(idx = -1) "c"
The reduce() function returns a new dataset which is a subset of the original
cout << endl << ">> d1 has only columns x,c" << endl;
std::unique_ptr<RooAbsData> d1{d.reduce({x, c})};
d1->Print("v");
cout << endl << ">> d2 has only column y" << endl;
std::unique_ptr<RooAbsData> d2{d.reduce({y})};
d2->Print("v");
cout << endl << ">> d3 has only the points with y>5.17" << endl;
std::unique_ptr<RooAbsData> d3{d.reduce("y>5.17")};
d3->Print("v");
cout << endl << ">> d4 has only columns x,c for data points with y>5.17" << endl;
std::unique_ptr<RooAbsData> d4{d.reduce({x, c}, "y>5.17")};
d4->Print("v");
>> d1 has only columns x,c DataStore d (d) Contains 1000 entries Observables: 1) x = 9 L(-10 - 10) "x" 2) c = Plus(idx = 1) "c" >> d2 has only column y DataStore d (d) Contains 1000 entries Observables: 1) y = 31.607 L(0 - 40) "y" >> d3 has only the points with y>5.17 [#1] INFO:InputArguments -- The formula y>5.17 claims to use the variables (x,y,c) but only (y) seem to be in use. inputs: y>5.17 DataStore d (d) Contains 973 entries Observables: 1) x = 9 L(-10 - 10) "x" 2) y = 31.607 L(0 - 40) "y" 3) c = Plus(idx = 1) "c" >> d4 has only columns x,c for data points with y>5.17 DataStore d (d) Contains 973 entries Observables: 1) x = 9 L(-10 - 10) "x" 2) c = Plus(idx = 1) "c"
The merge() function adds two data set column-wise
cout << endl << ">> merge d2(y) with d1(x,c) to form d1(x,c,y)" << endl;
static_cast<RooDataSet&>(*d1).merge(&static_cast<RooDataSet&>(*d2));
d1->Print("v");
>> merge d2(y) with d1(x,c) to form d1(x,c,y) DataStore d (d) Contains 1000 entries Observables: 1) x = 9 L(-10 - 10) "x" 2) c = Plus(idx = 1) "c" 3) y = 31.607 L(0 - 40) "y"
The append() function adds two datasets row-wise
cout << endl << ">> append data points of d3 to d1" << endl;
static_cast<RooDataSet&>(*d1).append(static_cast<RooDataSet&>(*d3));
d1->Print("v");
>> append data points of d3 to d1 DataStore d (d) Contains 1973 entries Observables: 1) x = 9 L(-10 - 10) "x" 2) c = Plus(idx = 1) "c" 3) y = 31.607 L(0 - 40) "y"
A binned dataset can be constructed empty, from an unbinned dataset, or from a ROOT native histogram (TH1,2,3)
cout << ">> construct dh (binned) from d(unbinned) but only take the x and y dimensions," << endl
<< ">> the category 'c' will be projected in the filling process" << endl;
>> construct dh (binned) from d(unbinned) but only take the x and y dimensions, >> the category 'c' will be projected in the filling process
The binning of real variables (like x,y) is done using their fit range 'get/setRange()' and number of specified fit bins 'get/setBins()'. Category dimensions of binned datasets get one bin per defined category state
x.setBins(10);
y.setBins(10);
RooDataHist dh("dh", "binned version of d", RooArgSet(x, y), d);
dh.Print("v");
RooPlot *yframe = y.frame(Bins(10), Title("Operations on binned datasets"));
dh.plotOn(yframe); // plot projection of 2D binned data on y
DataStore dh (binned version of d) Contains 100 entries Observables: 1) x = 9 L(-10 - 10) B(10) "x" 2) y = 38 L(0 - 40) B(10) "y" Binned Dataset dh (binned version of d) Contains 100 bins with a total weight of 1000 Observables: 1) x = 9 L(-10 - 10) B(10) "x" 2) y = 38 L(0 - 40) B(10) "y"
Examine the statistics of a binned dataset
cout << ">> number of bins in dh : " << dh.numEntries() << endl;
cout << ">> sum of weights in dh : " << dh.sum(false) << endl;
cout << ">> integral over histogram: " << dh.sum(true) << endl; // accounts for bin volume
>> number of bins in dh : 100 >> sum of weights in dh : 1000 >> integral over histogram: 8000
Locate a bin from a set of coordinates and retrieve its properties
x = 0.3;
y = 20.5;
cout << ">> retrieving the properties of the bin enclosing coordinate (x,y) = (0.3,20.5) " << endl;
cout << " bin center:" << endl;
dh.get(RooArgSet(x, y))->Print("v"); // load bin center coordinates in internal buffer
cout << " weight = " << dh.weight() << endl; // return weight of last loaded coordinates
>> retrieving the properties of the bin enclosing coordinate (x,y) = (0.3,20.5) bin center: 1) 0x7fdde962e2d0 RooRealVar:: x = 1 L(-10 - 10) B(10) "x" 2) 0x7fdde9a504c0 RooRealVar:: y = 22 L(0 - 40) B(10) "y" weight = 76
Reduce the 2-dimensional binned dataset to a 1-dimensional binned dataset
All reduce() methods are interfaced in RooAbsData. All reduction techniques demonstrated on unbinned datasets can be applied to binned datasets as well.
cout << ">> Creating 1-dimensional projection on y of dh for bins with x>0" << endl;
std::unique_ptr<RooAbsData> dh2{dh.reduce(y, "x>0")};
dh2->Print("v");
>> Creating 1-dimensional projection on y of dh for bins with x>0 DataStore dh (binned version of d) Contains 10 entries Observables: 1) y = 38 L(0 - 40) B(10) "y" Binned Dataset dh (binned version of d) Contains 10 bins with a total weight of 500 Observables: 1) y = 38 L(0 - 40) B(10) "y"
Add dh2 to yframe and redraw
dh2->plotOn(yframe, LineColor(kRed), MarkerColor(kRed));
[#1] INFO:Plotting -- RooPlot::updateFitRangeNorm: New event count of 500 will supersede previous event count of 1000 for normalization of PDF projections
Datasets can be persisted with ROOT I/O
cout << endl << ">> Persisting d via ROOT I/O" << endl;
TFile f("rf402_datahandling.root", "RECREATE");
d.Write();
f.ls();
>> Persisting d via ROOT I/O TFile** rf402_datahandling.root TFile* rf402_datahandling.root KEY: RooDataSet d;1 d KEY: TProcessID ProcessID0;1 225bfb26-e625-11ee-95d6-942c8a89beef
To read back in future session:
TFile f("rf402_datahandling.root") ; RooDataSet* d = (RooDataSet*) f.FindObject("d") ;
new TCanvas("rf402_datahandling", "rf402_datahandling", 600, 600);
gPad->SetLeftMargin(0.15);
yframe->GetYaxis()->SetTitleOffset(1.4);
yframe->Draw();
Draw all canvases
%jsroot on
gROOT->GetListOfCanvases()->Draw()