HybridOriginalDemo.C

Go to the documentation of this file.

// Example on how to use the HybridCalculatorOriginal class
//
// Author: Gregory Schott
#include "RooRandom.h"
#include "RooRealVar.h"
#include "RooGaussian.h"
#include "RooPolynomial.h"
#include "RooArgSet.h"
#include "RooAddPdf.h"
#include "RooDataSet.h"
#include "RooExtendPdf.h"
#include "RooConstVar.h"

#ifndef __CINT__  // problem including this file with CINT
#include "RooGlobalFunc.h"
#endif

#include "RooStats/HybridCalculatorOriginal.h"
#include "RooStats/HybridResult.h"
#include "RooStats/HybridPlot.h"

void HybridOriginalDemo(int ntoys = 1000)
{
  //***********************************************************************//
  // This macro show an example on how to use RooStats/HybridCalculatorOriginal //
  //***********************************************************************//
  //
  // With this example, you should get: CL_sb = 0.130 and CL_b = 0.946
  // (if data had -2lnQ = -3.0742). You can compare to the expected plot:
  // http://www-ekp.physik.uni-karlsruhe.de/~schott/roostats/hybridplot_example.png

  using namespace RooFit;
  using namespace RooStats;

  /// set RooFit random seed
  RooRandom::randomGenerator()->SetSeed(3007);

  /// build the models for background and signal+background
  RooRealVar x("x","",-3,3);
  RooArgList observables(x); // variables to be generated

  // gaussian signal
//  RooRealVar sig_mean("sig_mean","",0);
//  RooRealVar sig_sigma("sig_sigma","",0.8);
//  RooGaussian sig_pdf("sig_pdf","",x,sig_mean,sig_sigma);
  RooGaussian sig_pdf("sig_pdf","",x, RooConst(0.0),RooConst(0.8));
  RooRealVar sig_yield("sig_yield","",20,0,300);

  // flat background (extended PDF)
//  RooRealVar bkg_slope("bkg_slope","",0);
//  RooPolynomial bkg_pdf("bkg_pdf","",x,bkg_slope);
  RooPolynomial bkg_pdf("bkg_pdf","", x, RooConst(0));
  RooRealVar bkg_yield("bkg_yield","",40,0,300);
  RooExtendPdf bkg_ext_pdf("bkg_ext_pdf","",bkg_pdf,bkg_yield);

//  bkg_yield.setConstant(kTRUE);
  sig_yield.setConstant(kTRUE);

  // total sig+bkg (extended PDF)
  RooAddPdf tot_pdf("tot_pdf","",RooArgList(sig_pdf,bkg_pdf),RooArgList(sig_yield,bkg_yield));

  /// build the prior PDF on the parameters to be integrated
  // gaussian contraint on the background yield ( N_B = 40 +/- 10  ie. 25% )
  RooGaussian bkg_yield_prior("bkg_yield_prior","",bkg_yield,RooConst(bkg_yield.getVal()),RooConst(10.));

  RooArgSet nuisance_parameters(bkg_yield); // variables to be integrated

  /// generate a data sample
  RooDataSet* data = tot_pdf.generate(observables,RooFit::Extended());

  //***********************************************************************//

  /// run HybridCalculator on those inputs

  // use interface from HypoTest calculator by default

  HybridCalculatorOriginal myHybridCalc(*data, tot_pdf , bkg_ext_pdf ,
                                   &nuisance_parameters, &bkg_yield_prior);

  // here I use the default test statistics: 2*lnQ (optional)
  myHybridCalc.SetTestStatistic(1);
  //myHybridCalc.SetTestStatistic(3); // profile likelihood ratio

  myHybridCalc.SetNumberOfToys(ntoys);
  myHybridCalc.UseNuisance(true);

  // for speed up generation (do binned data)
  myHybridCalc.SetGenerateBinned(false);

  // calculate by running ntoys for the S+B and B hypothesis and retrieve the result
  HybridResult* myHybridResult = myHybridCalc.GetHypoTest();

  if (! myHybridResult) {
     std::cerr << "\nError returned from Hypothesis test" << std::endl;
     return;
  }

  /// run 1000 toys without gaussian prior on the background yield
  //HybridResult* myHybridResult = myHybridCalc.Calculate(*data,1000,false);

  /// save the toy-MC results to file, this way splitting into sub-batch jobs is possible
  //TFile fileOut("some_hybridresult.root","RECREATE");
  //fileOut.cd();
  //myHybridResult.Write();
  //fileOut.Close();

  /// read the results from a file
  //TFile fileIn("some_hybridresult.root");
  //HybridResult* myOtherHybridResult = (HybridResult*) fileIn.Get("myHybridCalc");

  /// example on how to merge with toy-MC results obtained in another job
  //HybridResult* mergedHybridResult = new HybridResult("mergedHybridResult","this object holds merged results");
  //mergedHybridResult->Add(myHybridResult);
  //mergedHybridResult->Add(myOtherHybridResult);
  /// or
  //myHybridResult->Add(myOtherHybridResult);

  /// nice plot of the results
  HybridPlot* myHybridPlot = myHybridResult->GetPlot("myHybridPlot","Plot of results with HybridCalculatorOriginal",100);
  myHybridPlot->Draw();

  /// recover and display the results
  double clsb_data = myHybridResult->CLsplusb();
  double clb_data = myHybridResult->CLb();
  double cls_data = myHybridResult->CLs();
  double data_significance = myHybridResult->Significance();
  double min2lnQ_data = myHybridResult->GetTestStat_data();

  /// compute the mean expected significance from toys
  double mean_sb_toys_test_stat = myHybridPlot->GetSBmean();
  myHybridResult->SetDataTestStatistics(mean_sb_toys_test_stat);
  double toys_significance = myHybridResult->Significance();

  std::cout << "Completed HybridCalculatorOriginal example:\n";
  std::cout << " - -2lnQ = " << min2lnQ_data << endl;
  std::cout << " - CL_sb = " << clsb_data << std::endl;
  std::cout << " - CL_b  = " << clb_data << std::endl;
  std::cout << " - CL_s  = " << cls_data << std::endl;
  std::cout << " - significance of data  = " << data_significance << std::endl;
  std::cout << " - mean significance of toys  = " << toys_significance << std::endl;
}