HybridResult.cxx

Go to the documentation of this file.

// @(#)root/roostats:$Id$

/*************************************************************************
 * Project: RooStats                                                     *
 * Package: RooFit/RooStats                                              *
 * Authors:                                                              *
 *   Kyle Cranmer, Lorenzo Moneta, Gregory Schott, Wouter Verkerke       *
 *************************************************************************
 * Copyright (C) 1995-2008, Rene Brun and Fons Rademakers.               *
 * All rights reserved.                                                  *
 *                                                                       *
 * For the licensing terms see $ROOTSYS/LICENSE.                         *
 * For the list of contributors see $ROOTSYS/README/CREDITS.             *
 *************************************************************************/

/** \class RooStats::HybridResult
    \ingroup Roostats

Class encapsulating the result of the HybridCalculatorOriginal.
This class is a fresh rewrite in RooStats of
RooStatsCms/LimitResults developed by D. Piparo and G. Schott
New contributions to this class have been written by Matthias Wolf (error estimation)

The objects of this class store and access with lightweight methods the
information calculated by LimitResults through a Lent calculation using
MC toy experiments.
In some ways can be considered an extended and extensible implementation of the
TConfidenceLevel class (http://root.cern.ch/root/html/TConfidenceLevel.html).

*/


#include "RooDataHist.h"
#include "RooDataSet.h"
#include "RooGlobalFunc.h" // for RooFit::Extended()
#include "RooNLLVar.h"
#include "RooRealVar.h"
#include "RooAbsData.h"

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


/// ClassImp for building the THtml documentation of the class
using namespace std;

ClassImp(RooStats::HybridResult);

using namespace RooStats;

////////////////////////////////////////////////////////////////////////////////
/// Construtor

HybridResult::HybridResult( const char *name) :
   HypoTestResult(name),
   fTestStat_data(-999.),
   fComputationsNulDoneFlag(false),
   fComputationsAltDoneFlag(false),
   fSumLargerValues(false)
{
   // HybridResult default constructor (with name )
}

////////////////////////////////////////////////////////////////////////////////
/// Construtor

HybridResult::HybridResult( const char *name,
                            const std::vector<double>& testStat_sb_vals,
                            const std::vector<double>& testStat_b_vals,
             bool sumLargerValues ) :
   HypoTestResult(name,0,0),
   fTestStat_data(-999.),
   fComputationsNulDoneFlag(false),
   fComputationsAltDoneFlag(false),
   fSumLargerValues(sumLargerValues)
{
   // HybridResult constructor (with name, title and vectors of S+B and B values)

   int vector_size_sb = testStat_sb_vals.size();
   assert(vector_size_sb>0);

   int vector_size_b = testStat_b_vals.size();
   assert(vector_size_b>0);

   fTestStat_sb.reserve(vector_size_sb);
   for (int i=0;i<vector_size_sb;++i)
      fTestStat_sb.push_back(testStat_sb_vals[i]);

   fTestStat_b.reserve(vector_size_b);
   for (int i=0;i<vector_size_b;++i)
      fTestStat_b.push_back(testStat_b_vals[i]);
}


////////////////////////////////////////////////////////////////////////////////
/// HybridResult destructor

HybridResult::~HybridResult()
{

   fTestStat_sb.clear();
   fTestStat_b.clear();
}

////////////////////////////////////////////////////////////////////////////////
/// set the value of the test statistics on data

void HybridResult::SetDataTestStatistics(double testStat_data_val)
{
   fComputationsAltDoneFlag = false;
   fComputationsNulDoneFlag = false;
   fTestStat_data = testStat_data_val;
   return;
}

////////////////////////////////////////////////////////////////////////////////
/// Returns \f$1 - CL_{b}\f$ : the B p-value

double HybridResult::NullPValue() const
{
   if (fComputationsNulDoneFlag==false) {
      int nToys = fTestStat_b.size();
      if (nToys==0) {
         std::cout << "Error: no toy data present. Returning -1.\n";
         return -1;
      }

      double larger_than_measured=0;
      if (fSumLargerValues) {
   for (int iToy=0;iToy<nToys;++iToy)
     if ( fTestStat_b[iToy] >= fTestStat_data ) ++larger_than_measured;
      } else {
   for (int iToy=0;iToy<nToys;++iToy)
     if ( fTestStat_b[iToy] <= fTestStat_data ) ++larger_than_measured;
      }

      if (larger_than_measured==0) std::cout << "Warning: CLb = 0 ... maybe more toys are needed!\n";

      fComputationsNulDoneFlag = true;
      fNullPValue = 1-larger_than_measured/nToys;
   }

   return fNullPValue;
}

////////////////////////////////////////////////////////////////////////////////
/// Returns \f$CL_{s+b}\f$ : the S+B p-value

double HybridResult::AlternatePValue() const
{
   if (fComputationsAltDoneFlag==false) {
      int nToys = fTestStat_b.size();
      if (nToys==0) {
         std::cout << "Error: no toy data present. Returning -1.\n";
         return -1;
      }

      double larger_than_measured=0;
      if (fSumLargerValues) {
   for (int iToy=0;iToy<nToys;++iToy)
     if ( fTestStat_sb[iToy] >= fTestStat_data ) ++larger_than_measured;
      } else {
   for (int iToy=0;iToy<nToys;++iToy)
     if ( fTestStat_sb[iToy] <= fTestStat_data ) ++larger_than_measured;
      }

      if (larger_than_measured==0) std::cout << "Warning: CLsb = 0 ... maybe more toys are needed!\n";

      fComputationsAltDoneFlag = true;
      fAlternatePValue = larger_than_measured/nToys;
   }

   return fAlternatePValue;
}

////////////////////////////////////////////////////////////////////////////////
/// Returns an estimate of the error on \f$CL_{b}\f$ assuming a binomial
/// error on \f$CL_{b}\f$:
/// \f[
/// \sigma_{CL_{b}} = \sqrt{CL_{b} \left( 1 - CL_{b} \right) / n_{toys}}
/// \f]

Double_t HybridResult::CLbError() const
{
  unsigned const int n = fTestStat_b.size();
  return TMath::Sqrt(CLb() * (1. - CLb()) / n);
}

////////////////////////////////////////////////////////////////////////////////
/// Returns an estimate of the error on \f$CL_{s+b}\f$ assuming a binomial
/// error on \f$CL_{s+b}\f$:
/// \f[
/// \sigma_{CL_{s+b}} = \sqrt{CL_{s+b} \left( 1 - CL_{s+b} \right) / n_{toys}}
/// \f]

Double_t HybridResult::CLsplusbError() const
{
  unsigned const int n = fTestStat_sb.size();
  return TMath::Sqrt(CLsplusb() * (1. - CLsplusb()) / n);
}

////////////////////////////////////////////////////////////////////////////////
/// Returns an estimate of the error on \f$CL_{s}\f$ through combination
/// of the errors on \f$CL_{b}\f$ and \f$CL_{s+b}\f$:
/// \f[
/// \sigma_{CL_s} = CL_s \sqrt{\left( \frac{\sigma_{CL_{s+b}}}{CL_{s+b}} \right)^2 + \left( \frac{\sigma_{CL_{b}}}{CL_{b}} \right)^2}
/// \f]

Double_t HybridResult::CLsError() const
{
  unsigned const int n_b = fTestStat_b.size();
  unsigned const int n_sb = fTestStat_sb.size();

  if (CLb() == 0 || CLsplusb() == 0)
    return 0;

  double cl_b_err = (1. - CLb()) / (n_b * CLb());
  double cl_sb_err = (1. - CLsplusb()) / (n_sb * CLsplusb());

  return CLs() * TMath::Sqrt(cl_b_err + cl_sb_err);
}

////////////////////////////////////////////////////////////////////////////////
/// add additional toy-MC experiments to the current results
/// use the data test statistics of the added object if none is already present
/// (otherwise, ignore the new one)

void HybridResult::Add(HybridResult* other)
{

   int other_size_sb = other->GetTestStat_sb().size();
   for (int i=0;i<other_size_sb;++i)
      fTestStat_sb.push_back(other->GetTestStat_sb()[i]);

   int other_size_b = other->GetTestStat_b().size();
   for (int i=0;i<other_size_b;++i)
      fTestStat_b.push_back(other->GetTestStat_b()[i]);

   // if no data is present use the other's HybridResult's data
   if (fTestStat_data==-999.)
      fTestStat_data = other->GetTestStat_data();

   fComputationsAltDoneFlag = false;
   fComputationsNulDoneFlag = false;

   return;
}

////////////////////////////////////////////////////////////////////////////////
/// prepare a plot showing a result and return a pointer to a HybridPlot object
/// the needed arguments are: an object name, a title and the number of bins in the plot

HybridPlot* HybridResult::GetPlot(const char* name,const char* title, int n_bins)
{
   // default plot name
   TString plot_name;
   if ( TString(name)=="" ) {
      plot_name += GetName();
      plot_name += "_plot";
   } else plot_name = name;

   // default plot title
   TString plot_title;
   if ( TString(title)=="" ) {
      plot_title += GetTitle();
      plot_title += "_plot (";
      plot_title += fTestStat_b.size();
      plot_title += " toys)";
   } else plot_title = title;

   HybridPlot* plot = new HybridPlot( plot_name.Data(),
                                      plot_title.Data(),
                                      fTestStat_sb,
                                      fTestStat_b,
                                      fTestStat_data,
                                      n_bins,
                                      true );
   return plot;
}

////////////////////////////////////////////////////////////////////////////////
/// Print out some information about the results

void HybridResult::PrintMore(const char* /*options */)
{
   std::cout << "\nResults " << GetName() << ":\n"
             << " - Number of S+B toys: " << fTestStat_b.size() << std::endl
             << " - Number of B toys: " << fTestStat_sb.size() << std::endl
             << " - test statistics evaluated on data: " << fTestStat_data << std::endl
             << " - CL_b " << CLb() << std::endl
             << " - CL_s+b " << CLsplusb() << std::endl
             << " - CL_s " << CLs() << std::endl;

   return;
}