APWeightSum Class Reference

Class to calculate the sum of weights ("weighted counter") More...

#include <APWeightSum.h>

Inheritance diagram for APWeightSum:

Collaboration diagram for APWeightSum:

Public Member Functions
	APWeightSum ()
	Default constructor.
virtual	~APWeightSum ()
	Default destructor.
void	AddWeightToEvt (APWeightEntry *weight)
	Adds a weight to the sum of weights.
void	FinishEvt (double ext_weight=1.0)
	Finishes the current event and calculates the event weight.
void	AddEvt (APEvtWeight *evt_weight, double ext_weight=1.0)
	Adds an event with an externally calculated EvtWeight object.
double	GetSumW () const
	Returns the sum of weights.
double	GetSumW2 () const
	Returns sum of (weights^2).
double	GetSumWExternal () const
	Returns the sum of weights without taking into account the trigger weighting (external weights only) to allow switching trigger weighting on/off.
double	GetStdDev ()
	Returns the standard deviation.
double	GetVariance ()
	Returns the variance.
double	GetVarianceNoCorr ()
	Returns the variance, assuming no correlations.
double	GetVarianceFullCorr ()
	Returns the variance, assuming full correlation amongst objects.
double	GetSysUncert () const
	Returns the systematic uncertainty (from systematics assigned to weights).
unsigned long	GetKUnweighted () const
	Returns the unweighted sum of entries.
THnSparse *	GetUncertHistogram (APReweightBase *weighter)
	Returns THnSparse holding the uncertainties for given APReweightBase instance.
const std::vector< THnSparse * > &	GetAllUncertHistograms ()
	Returns vector of THnSparses holding the uncertainties for all APReweight IDs.

Public Attributes
ClassDef(APWeightSum, 1) protected std::vector< APWeightEntry * >	m_current_evt_weights
	< Calculates the final uncertainties including correlations.
std::vector< THnSparse * >	m_linear_uncert
	Holds all histograms for uncertainties.
unsigned long int	m_k_evt_orig
	Holds the original amount of unweighted counts ("sum of 1's").
double	m_k_evt_weight
	Holds the sum of weights.
double	m_k_evt_weight2
	Holds the sum of squared weights.
double	m_k_evt_weight_external
	Holds the sum of external weights (no trigger weighting).
double	m_variance
	Holds the variance.
double	m_variance_nocorr
	Holds the variance, assuming no correlations.
double	m_variance_fullcorr
	Holds the variance, assuming full correlation amongst objects.
double	m_variance_sys
	Holds the systematic variance (from systematics assigned to weights).
bool	m_isComputed

Detailed Description

Class to calculate the sum of weights ("weighted counter")

Calculates the sum of weights taking into account the underlying asymmetric probability distribution. This is done by modelling the pdf and then extracting the corresponding quantiles.

Author

fabia.nosp@m.n.Ko.nosp@m.hn@ce.nosp@m.rn.c.nosp@m.h

Definition at line 29 of file APWeightSum.h.

Constructor & Destructor Documentation

APWeightSum()

APWeightSum::APWeightSum

(

)

Default constructor.

Definition at line 17 of file APWeightSum.cxx.

                         {
  m_k_evt_orig = 0;
  m_k_evt_weight = 0.;
  m_k_evt_weight2 = 0.;
  m_k_evt_weight_external = 0.;
  m_variance = 0.;
  m_variance_nocorr = 0.;
  m_variance_fullcorr = 0.;
  m_variance_sys = 0.;
 
  m_isComputed = false;
}

~APWeightSum()

APWeightSum::~APWeightSum ( )

virtual

Default destructor.

Definition at line 30 of file APWeightSum.cxx.

                          {
  for (vector<THnSparse*>::reverse_iterator it = m_linear_uncert.rbegin(); it != m_linear_uncert.rend(); ++it) delete *it;
  m_linear_uncert.clear();
}

Member Function Documentation

AddEvt()

void APWeightSum::AddEvt	(	APEvtWeight *	evt_weight,
		double	ext_weight = 1.0 )

Adds an event with an externally calculated EvtWeight object.

Takes an optional external weight factor (e.g. mc weight).

Definition at line 100 of file APWeightSum.cxx.

                                                                   {
  ++m_k_evt_orig;
  m_k_evt_weight += ext_weight * evt_weight->GetWeight();
  m_k_evt_weight2 += m_k_evt_weight*m_k_evt_weight;
  m_k_evt_weight_external += ext_weight;
 
  vector<APWeightEntry*> temp_vec_mu = evt_weight->GetWeightObjects(APEvtWeight::kMuon);
  vector<APWeightEntry*> temp_vec_dimu = evt_weight->GetWeightObjects(APEvtWeight::kDiMuon);
  vector<APWeightEntry*> temp_vec_mumo = evt_weight->GetWeightObjects(APEvtWeight::kMuonMO);
  vector<APWeightEntry*> temp_vec_tau = evt_weight->GetWeightObjects(APEvtWeight::kTau);
  vector<APWeightEntry*> temp_vec_ditau = evt_weight->GetWeightObjects(APEvtWeight::kDiTau);
  vector<APWeightEntry*> temp_vec_taumo = evt_weight->GetWeightObjects(APEvtWeight::kTauMO);
  vector<APWeightEntry*> temp_vec_el = evt_weight->GetWeightObjects(APEvtWeight::kElectron);
  vector<APWeightEntry*> temp_vec_diel = evt_weight->GetWeightObjects(APEvtWeight::kDiElectron);
  vector<APWeightEntry*> temp_vec_elmo = evt_weight->GetWeightObjects(APEvtWeight::kElectronMO);
  vector<APWeightEntry*> temp_vec_jet = evt_weight->GetWeightObjects(APEvtWeight::kJet);
  vector<APWeightEntry*> temp_vec_dijet = evt_weight->GetWeightObjects(APEvtWeight::kDiJet);
  vector<APWeightEntry*> temp_vec_jetmo = evt_weight->GetWeightObjects(APEvtWeight::kJetMO);
  
  vector< vector<APWeightEntry*> > temp_vec_all{
    temp_vec_mu, temp_vec_tau, temp_vec_el, temp_vec_jet,
    std::move(temp_vec_mumo), std::move(temp_vec_taumo), std::move(temp_vec_elmo), std::move(temp_vec_jetmo),
    temp_vec_dimu, temp_vec_ditau, temp_vec_diel, temp_vec_dijet
  };
  
 
  /* check if histogram for error propagation is already there; if not, create it */
  for( unsigned int iAll = 0, IAll = temp_vec_all.size(); iAll < IAll; ++iAll ) {
    for( unsigned int i = 0, I = temp_vec_all[iAll].size(); i < I; ++i ) {
      unsigned int ID = temp_vec_all[iAll][i]->GetID();
      if( m_linear_uncert.size() < ID+1 ) m_linear_uncert.resize(ID+1, 0);
      if( m_linear_uncert[ID] == 0 ) {
        vector<int> original_dimensions = temp_vec_all[iAll][i]->GetOriginalDimensions();
        int *bins = new int[original_dimensions.size()];
        double *xmin = new double[original_dimensions.size()];
        double *xmax = new double[original_dimensions.size()];
        for( unsigned int j = 0, J = original_dimensions.size(); j < J; ++j ) {
          bins[j] = original_dimensions[j];
          xmin[j] = 0.;
          xmax[j] = 10.;
        }
        m_linear_uncert[ID] = new THnSparseD("","",original_dimensions.size(), bins, xmin, xmax);
      }
    }
  }
  /* calculate weight and derivatives */
  /* these are the simple weights: kMuon, kTau, kElectron and kJet: single object trigger, OR of all elements */
  if(evt_weight->GetType() <= APEvtWeight::kJet) {
    vector<APWeightEntry*> temp_vec_rel;
    if(evt_weight->GetType() == APEvtWeight::kMuon) temp_vec_rel = std::move(temp_vec_mu);
    else if(evt_weight->GetType() == APEvtWeight::kTau) temp_vec_rel = std::move(temp_vec_tau);
    else if(evt_weight->GetType() == APEvtWeight::kElectron) temp_vec_rel = std::move(temp_vec_el);
    else if(evt_weight->GetType() == APEvtWeight::kJet) temp_vec_rel = std::move(temp_vec_jet);
    
    for (unsigned int i = 0, I = temp_vec_rel.size(); i < I; ++i ) {
      vector<int> coord = temp_vec_rel[i]->GetCoords();
      double weight_uncert = sqrt(temp_vec_rel[i]->GetVariance());
      for (unsigned int j = 0; j < I; ++j ) {
        if (j == i) continue;
        weight_uncert *= (1.0 - temp_vec_rel[j]->GetExpectancy());
      }
      m_linear_uncert[temp_vec_rel[i]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_rel[i]->GetID()]->GetBinContent(&coord.front())+weight_uncert);
      m_variance_nocorr += weight_uncert*weight_uncert;
    }
  }
  
  /* these are the DiMuon, DiTau, DiElectron and DiJet weights */
  else if (evt_weight->GetType() >= APEvtWeight::kDiMuon && evt_weight->GetType() <= APEvtWeight::kDiJet) {
    vector<APWeightEntry*> temp_vec_rel;
    if(evt_weight->GetType() == APEvtWeight::kDiMuon) temp_vec_rel = std::move(temp_vec_dimu);
    else if(evt_weight->GetType() == APEvtWeight::kDiTau) temp_vec_rel = std::move(temp_vec_ditau);
    else if(evt_weight->GetType() == APEvtWeight::kDiElectron) temp_vec_rel = std::move(temp_vec_diel);
    else if(evt_weight->GetType() == APEvtWeight::kDiJet) temp_vec_rel = std::move(temp_vec_dijet);
    
    bool isAsymTrig = false;
    vector<unsigned int> temp_vec_IDs;
    temp_vec_IDs.push_back(temp_vec_rel[0]->GetID() );
    for( unsigned int i = 1, I = temp_vec_rel.size(); i < I; ++i ) {
      bool knownID = false;
      for( unsigned int j = 0, J = temp_vec_IDs.size(); j < J; ++j ) {
        if( temp_vec_rel[i]->GetID() == temp_vec_IDs[j] ) { knownID = true; break; }
      }
      if( !knownID ) temp_vec_IDs.push_back( temp_vec_rel[i]->GetID() );
    }
    if( temp_vec_IDs.size() != 1 ) isAsymTrig = true;
 
    /* this is for symmetric dilepton triggers */
    if( !isAsymTrig ) {
      for (unsigned int i = 0, I = temp_vec_rel.size(); i < I; ++i ) {
        vector<int> coord = temp_vec_rel[i]->GetCoords();
        double weight_uncert = sqrt(temp_vec_rel[i]->GetVariance());
        double weight_derivative = 0.;
        for (unsigned int j = 0; j < I; ++j ) {
          if (j == i) continue;
          double weight_derivative_temp = temp_vec_rel[j]->GetExpectancy();
          for (unsigned int k = 0; k < I; ++k ) {
            if( k == j || k == i ) continue;
            weight_derivative_temp *= (1.0 - temp_vec_rel[k]->GetExpectancy());
          }
          weight_derivative += weight_derivative_temp;
        }
        weight_uncert *= weight_derivative;
        m_linear_uncert[temp_vec_rel[i]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_rel[i]->GetID()]->GetBinContent(&coord.front())+weight_uncert);
        m_variance_nocorr += weight_uncert*weight_uncert;
      }
    }
    
    /* this is for asymmetric triggers */
    else {
      /* at first the first leg of the trigger */ 
      for( unsigned int k = 0, K = temp_vec_rel.size(); k < K; k += 4 ) {
        vector<int> coord = temp_vec_rel[k]->GetCoords();
        double variance_k = 0.;
        double variance_temp = 1.;
        for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
          if( i == k ) continue;
          variance_temp *= (1. - temp_vec_rel[i]->GetExpectancy());
        }
        variance_k += variance_temp;
        variance_temp = -1.;
        for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
          variance_temp *= (1.0 - temp_vec_rel[i+2]->GetExpectancy());
          if( i == k ) continue;
          variance_temp *= (1.0 - temp_vec_rel[i]->GetExpectancy());
        }
        variance_k += variance_temp;
        variance_temp = -temp_vec_rel[k+3]->GetExpectancy();
        for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
          if( i == k ) continue;
          variance_temp *= (1.0 - temp_vec_rel[i]->GetExpectancy())*(1.0 - temp_vec_rel[i+2]->GetExpectancy());
        }
        variance_k += variance_temp;
        variance_temp = 0.;
        for( unsigned int j = 0, J = temp_vec_rel.size(); j < J; j+= 4 ) {
          if( j == k ) continue;
          double variance_ijk_temp = temp_vec_rel[j]->GetExpectancy()*temp_vec_rel[j+3]->GetExpectancy();
          for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
            if( i == j ) continue;
            variance_ijk_temp *= (1.0 -  temp_vec_rel[i+2]->GetExpectancy());
            if( i == k ) continue;
            variance_ijk_temp *= (1.0 - temp_vec_rel[i]->GetExpectancy());
          }
          variance_temp += variance_ijk_temp;
        }
        variance_k += variance_temp;
        variance_k *= variance_k*temp_vec_rel[k]->GetVariance();
        m_linear_uncert[temp_vec_rel[k]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_rel[k]->GetID()]->GetBinContent(&coord.front())+sqrt(variance_k));
        m_variance_nocorr += variance_k;
      }
      
      /* second leg */
      for( unsigned int k = 0, K = temp_vec_rel.size(); k < K; k += 4 ) {
        vector<int> coord = temp_vec_rel[k+1]->GetCoords();
        double variance_temp = 1.;
        for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
        if( i == k ) continue;
          variance_temp *= (1. - temp_vec_rel[i+1]->GetExpectancy());
        }
        variance_temp *= variance_temp*temp_vec_rel[k+1]->GetVariance();
        m_linear_uncert[temp_vec_rel[k+1]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_rel[k+1]->GetID()]->GetBinContent(&coord.front())+sqrt(variance_temp));
        m_variance_nocorr += variance_temp;
      }
      
      /* second leg | condition leg 1 failed */
      for( unsigned int k = 0, K = temp_vec_rel.size(); k < K; k += 4 ) {
        vector<int> coord = temp_vec_rel[k+2]->GetCoords();
        double variance_k = 0.;
        double variance_temp = 1.;
        for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
          variance_temp *= (1. - temp_vec_rel[i]->GetExpectancy());
          if( i == k ) continue;
          variance_temp *= (1. - temp_vec_rel[i+2]->GetExpectancy());
        }
        variance_k += variance_temp;
        variance_temp = 0.;
        for( unsigned int j = 0, J = temp_vec_rel.size(); j < J; j+= 4 ) {
          double variance_ijk_temp = temp_vec_rel[j]->GetExpectancy()*temp_vec_rel[j+3]->GetExpectancy();
          for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
            if( j == i ) continue;
            variance_ijk_temp *= (1.0 - temp_vec_rel[i]->GetExpectancy());
            if( i == k ) continue;
            variance_ijk_temp *= (1.0 -  temp_vec_rel[i+2]->GetExpectancy());
          }
          variance_temp += variance_ijk_temp;
        }
        variance_k += variance_temp;
        variance_k *= variance_k*temp_vec_rel[k+2]->GetVariance();
        m_linear_uncert[temp_vec_rel[k+2]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_rel[k+2]->GetID()]->GetBinContent(&coord.front())+sqrt(variance_k));
        m_variance_nocorr += variance_k;
      }
      
      /* second leg | condition leg 1 passed */
      for( unsigned int k = 0, K = temp_vec_rel.size(); k < K; k += 4 ) {
        vector<int> coord = temp_vec_rel[k+3]->GetCoords();
        double variance_k = - temp_vec_rel[k]->GetExpectancy();
        for( unsigned int i = 0, I = temp_vec_rel.size(); i < I; i += 4 ) {
          if( i == k ) continue;
          variance_k *= (1.0 - temp_vec_rel[i]->GetExpectancy())*(1.0 - temp_vec_rel[i+3]->GetExpectancy());
        }
        variance_k *= variance_k*temp_vec_rel[k+3]->GetVariance();
        m_linear_uncert[temp_vec_rel[k+3]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_rel[k+3]->GetID()]->GetBinContent(&coord.front())+sqrt(variance_k));
        m_variance_nocorr += variance_k;
      }
    } 
    
  }
   
  /* these are the weights for ANDed triggers. Shouldn't be used but for examples. This will treat everything as ANDed. */
  else if (evt_weight->GetType() == APEvtWeight::kANDed ) {
    for (unsigned int iAll = 0, IAll = temp_vec_all.size(); iAll < IAll; ++iAll ) {
      for (unsigned int i = 0, I = temp_vec_all[iAll].size(); i < I; ++i ) {
        vector<int> coord = temp_vec_all[iAll][i]->GetCoords();
        double weight_uncert = sqrt(temp_vec_all[iAll][i]->GetVariance());
        weight_uncert *= (temp_vec_all[iAll][i]->GetExpectancy() <= numeric_limits<double>::epsilon() ) ? 0. : evt_weight->GetWeight()/temp_vec_all[iAll][i]->GetExpectancy();
        m_linear_uncert[temp_vec_all[iAll][i]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_all[iAll][i]->GetID()]->GetBinContent(&coord.front())+weight_uncert); 
        m_variance_nocorr += weight_uncert*weight_uncert;
      }
    }
  }
 
  /* these are weights for ORed triggers, for instance dimuon OR single muon */
  /* treat everything as ORed: calculate weight for each type, then combine weights */
  /* objects of same type are ORed: kMuon, kTau, kElectron, kJet, kMuonMO, kTauMO, kElectronMO, kJetMo */
  /* objects of different type are handled differently */
  /* if type is kMOANDed, everything is ANDed.*/
  /* if type is kMOORed, kMuonMO, kTauMO, kElectronMO, kJetMO are ANDed; all others are ORed with each other! */
  else if (evt_weight->GetType() >= APEvtWeight::kORed ) {
    
    vector<double> temp_weight_rel(12,1.);
    
    /* calculate single object trigger weights: OR of all objects of same type: kMuon, kTau, kElectron, kJet, kMuonMO, kTauMO, kElectronMO, kJetMO */
    for (unsigned int j = 0; j < 8; ++j) {
      for (unsigned int i = 0, I = temp_vec_all[j].size(); i < I; ++i ) temp_weight_rel[j] *= (1.0 - temp_vec_all[j][i]->GetExpectancy());
      // for triggers intended to be used in ORing, set weight to 1. if there is no object and an AND is require: (if no object is added, assume none is required)
      if( j < 4 ) temp_weight_rel[j] = (temp_vec_all[j].size() > 0 || evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed) ? (1.0 - temp_weight_rel[j]) : 1.;
      // for triggers inteded to be used in ANDing, set weight to 1. if there is no object added (if none is added assume none is required)
      else temp_weight_rel[j] = (temp_vec_all[j].size() > 0) ? (1.0 - temp_weight_rel[j]) : 1.;
    }
 
    
    /* calculate diobject trigger weights: OR of all objects of same type but require at least 2: kDiMuon, kDiTau, kDiElectron, kDiJet */
    for (unsigned int j = 8; j < 12; ++j) {
      if( temp_vec_all[j].size() >= 2 ) {
        for (unsigned int i = 0, I = temp_vec_all[j].size(); i < I; ++i) temp_weight_rel[j] *= (1.0 - temp_vec_all[j][i]->GetExpectancy());
        for (unsigned int i = 0, I = temp_vec_all[j].size(); i < I; ++i) {
          double temp_weight = temp_vec_all[j][i]->GetExpectancy();
          for (unsigned int k = 0; k < I; ++k ) {
            if( k == i ) continue;
            temp_weight *= (1.0 - temp_vec_all[j][k]->GetExpectancy());
          }
          temp_weight_rel[j] += temp_weight;
        }
      }
      temp_weight_rel[j] = (temp_vec_all[j].size() > 0 || evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed) ? (1.0 - temp_weight_rel[j]) : 1.;
    }
   
    /* calculat weight for multiobject trigger (AND of all MO objects) */
    double temp_weight_MO = 1.;
    int n_noObject_MO = 0;
    for (unsigned int l = 4; l < 8; ++l ) {
      temp_weight_MO *= temp_weight_rel[l];
      if( temp_vec_all[l].size() == 0 ) n_noObject_MO += 1;
    }
    if( /*(evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed) &&*/ n_noObject_MO >= 2 ) temp_weight_MO = 0.; 
 
    /* all partial weights calculated; now calculate uncertainties */
    /* single object triggers */
    for (unsigned int j = 0; j < 8; ++j) {
      for (unsigned int i = 0, I = temp_vec_all[j].size(); i < I; ++i ) {
        vector<int> coord = temp_vec_all[j][i]->GetCoords();
        double weight_uncert = sqrt(temp_vec_all[j][i]->GetVariance());
        for (unsigned int k = 0; k < I; ++k ) {
          if( k == i ) continue;
          weight_uncert *= (1.0 - temp_vec_all[j][k]->GetExpectancy());
        }
        for( unsigned int l = 0; l < 4; ++l ) {
          if( l == j ) continue;
          if( evt_weight->GetType() == APEvtWeight::kMOANDed ) weight_uncert *= temp_weight_rel[l];
          else if( evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed ) weight_uncert *= (1.0 - temp_weight_rel[l]); 
          else cout << "WARNING: handling for this weight type is unknown! uncertainties will be incorrect" << endl;
        }
        for( unsigned int l = 8; l < 12; ++l ) {
          if( l == j ) continue;
          if( evt_weight->GetType() == APEvtWeight::kMOANDed ) weight_uncert *= temp_weight_rel[l];
          else if( evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed ) weight_uncert *= (1.0 - temp_weight_rel[l]);
          else cout << "WARNING: handling for this weight type is unknown! uncertainties will be incorrect" << endl;
        }
 
        if( j < 4 || j > 7 ) {
          if( evt_weight->GetType() == APEvtWeight::kMOANDed ) weight_uncert *= temp_weight_MO;
          else if( evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed ) weight_uncert *= (1.0 - temp_weight_MO);
          else cout << "WARNING: handling for this weight type is unknown! uncertainties will be incorrect" << endl;
        }
        else if( j >= 4 && j <= 7 && temp_weight_rel[j] > numeric_limits<double>::epsilon() ) weight_uncert = weight_uncert*temp_weight_MO/temp_weight_rel[j];
        m_linear_uncert[temp_vec_all[j][i]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_all[j][i]->GetID()]->GetBinContent(&coord.front())+weight_uncert);
        m_variance_nocorr += weight_uncert*weight_uncert;
      }
    }
    
    /* diobject triggers */
    for (unsigned int j = 8; j < 12; ++j) {
      if( temp_vec_all[j].size() >= 2 ) {
        for (unsigned int i = 0, I = temp_vec_all[j].size(); i < I; ++i ) {
          vector<int> coord = temp_vec_all[j][i]->GetCoords();
          double weight_uncert = sqrt(temp_vec_all[j][i]->GetVariance());
          double weight_derivative = 0.;
          for (unsigned int k = 0; k < I; ++k ) {
            if( k == i ) continue;
            double weight_derivative_temp = temp_vec_all[j][k]->GetExpectancy();
            for (unsigned int l = 0; l < I; ++l ) {
              if( l == k || l == i ) continue;
              weight_derivative_temp *= (1.0 - temp_vec_all[j][l]->GetExpectancy());
            }
            weight_derivative += weight_derivative_temp;
          }
          weight_uncert *= weight_derivative;
          for( unsigned int l = 0; l < 4; ++l ) {
            if( l == j ) continue;
            if( evt_weight->GetType() == APEvtWeight::kMOANDed ) weight_uncert *= temp_weight_rel[l];
            else if( evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed ) weight_uncert *= (1.0 - temp_weight_rel[l]);
            else cout << "WARNING: handling for this weight type is unknown! uncertainties will be incorrect" << endl;
          }
          for( unsigned int l = 8; l < 12; ++l ) {
            if( l == j ) continue;
            if( evt_weight->GetType() == APEvtWeight::kMOANDed ) weight_uncert *= temp_weight_rel[l];
            else if( evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed ) weight_uncert *= (1.0 - temp_weight_rel[l]);
            else cout << "WARNING: handling for this weight type is unknown! uncertainties will be incorrect" << endl;
          }
          if( evt_weight->GetType() == APEvtWeight::kMOANDed ) weight_uncert *= temp_weight_MO;
          else if( evt_weight->GetType() == APEvtWeight::kORed || evt_weight->GetType() == APEvtWeight::kMOORed ) weight_uncert *= (1.0 - temp_weight_MO);
          else cout << "WARNING: handling for this weight type is unknown! uncertainties will be incorrect" << endl;
          
          if( j >= 4 && j <= 7 && temp_weight_rel[j] > numeric_limits<double>::epsilon() ) weight_uncert /= temp_weight_rel[j];
          
          m_linear_uncert[temp_vec_all[j][i]->GetID()]->SetBinContent(&coord.front(), m_linear_uncert[temp_vec_all[j][i]->GetID()]->GetBinContent(&coord.front())+weight_uncert);
          m_variance_nocorr += weight_uncert*weight_uncert; 
        }
      }
    }
  }
 
  m_variance_sys += ext_weight * evt_weight->GetSysVariance();
  
  m_isComputed = false;
}

AddWeightToEvt()

void APWeightSum::AddWeightToEvt

(

APWeightEntry *

weight

)

Adds a weight to the sum of weights.

Definition at line 35 of file APWeightSum.cxx.

                                                      {
  m_current_evt_weights.push_back(weight);
}

FinishEvt()

void APWeightSum::FinishEvt

(

double

ext_weight = 1.0

)

Finishes the current event and calculates the event weight.

Takes an optional external weight factor (e.g. mc weight).

Definition at line 78 of file APWeightSum.cxx.

                                             {
  ++m_k_evt_orig;
  double evt_weight = 1.0;
  double uncert = 0.0;
  double uncert_sys = 0.0;
  for (unsigned int i = 0, I = m_current_evt_weights.size(); i < I; ++i) {
    double uncert_summand = 1.0;
    evt_weight *= (1. - m_current_evt_weights[i]->GetExpectancy());
    for (unsigned int k = 0; k < I; ++k) {
      if (i != k) uncert_summand *= (1. - m_current_evt_weights[i]->GetExpectancy());
    }
    uncert += (uncert_summand * uncert_summand * m_current_evt_weights[i]->GetVariance());
    uncert_sys += (uncert_summand * uncert_summand * m_current_evt_weights[i]->GetSysUncert2());
  }
  m_k_evt_weight += ext_weight * (1. - evt_weight);
  m_k_evt_weight2 += m_k_evt_weight*m_k_evt_weight;
  m_k_evt_weight_external += ext_weight;
  m_variance += fabs(ext_weight) * uncert;
  m_variance_sys += ext_weight * uncert_sys;
  m_current_evt_weights.clear();
}

GetAllUncertHistograms()

const vector< THnSparse * > & APWeightSum::GetAllUncertHistograms

(

)

Returns vector of THnSparses holding the uncertainties for all APReweight IDs.

Definition at line 472 of file APWeightSum.cxx.

                                                                {
  return m_linear_uncert;
}

GetKUnweighted()

unsigned long APWeightSum::GetKUnweighted

(

)

const

Returns the unweighted sum of entries.

Definition at line 74 of file APWeightSum.cxx.

                                                {
  return m_k_evt_orig;
}

GetStdDev()

double APWeightSum::GetStdDev

(

)

Returns the standard deviation.

Definition at line 51 of file APWeightSum.cxx.

                              {
  if ( !m_isComputed ) Compute();
  return sqrt(m_variance);
}

GetSumW()

double APWeightSum::GetSumW

(

)

const

Returns the sum of weights.

Definition at line 39 of file APWeightSum.cxx.

                                  {
  return m_k_evt_weight;
}

GetSumW2()

double APWeightSum::GetSumW2

(

)

const

Returns sum of (weights^2).

Definition at line 43 of file APWeightSum.cxx.

                                   {
  return m_k_evt_weight2;
}

GetSumWExternal()

double APWeightSum::GetSumWExternal

(

)

const

Returns the sum of weights without taking into account the trigger weighting (external weights only) to allow switching trigger weighting on/off.

Definition at line 47 of file APWeightSum.cxx.

                                          {
  return m_k_evt_weight_external;
}

GetSysUncert()

double APWeightSum::GetSysUncert

(

)

const

Returns the systematic uncertainty (from systematics assigned to weights).

Definition at line 70 of file APWeightSum.cxx.

                                       {
  return sqrt(m_variance_sys);
}

GetUncertHistogram()

THnSparse * APWeightSum::GetUncertHistogram

(

APReweightBase *

weighter

)

Returns THnSparse holding the uncertainties for given APReweightBase instance.

Definition at line 463 of file APWeightSum.cxx.

                                                                     {
  unsigned int temp_ID = weighter->GetID();
  if( temp_ID > m_linear_uncert.size()-1 ) {
    cout << "WARNING: ID unknown. Returning 0-pointer!" << endl;
    return 0;
  }
  else return m_linear_uncert[temp_ID];
}

GetVariance()

double APWeightSum::GetVariance

(

)

Returns the variance.

Definition at line 56 of file APWeightSum.cxx.

                                {
  if ( !m_isComputed ) Compute();
  return m_variance;
}

GetVarianceFullCorr()

double APWeightSum::GetVarianceFullCorr

(

)

Returns the variance, assuming full correlation amongst objects.

Definition at line 65 of file APWeightSum.cxx.

                                        {
  if ( !m_isComputed ) Compute();
  return m_variance_fullcorr;
}

GetVarianceNoCorr()

double APWeightSum::GetVarianceNoCorr

(

)

Returns the variance, assuming no correlations.

Definition at line 61 of file APWeightSum.cxx.

                                      {
  return m_variance_nocorr;
}

Member Data Documentation

m_current_evt_weights

ClassDef (APWeightSum,1) protected std::vector< APWeightEntry* > APWeightSum::m_current_evt_weights

< Calculates the final uncertainties including correlations.

Holds the expectancy values of the weights in the current event.

Definition at line 57 of file APWeightSum.h.

m_isComputed

bool APWeightSum::m_isComputed

Definition at line 67 of file APWeightSum.h.

m_k_evt_orig

unsigned long int APWeightSum::m_k_evt_orig

Holds the original amount of unweighted counts ("sum of 1's").

Definition at line 59 of file APWeightSum.h.

m_k_evt_weight

double APWeightSum::m_k_evt_weight

Holds the sum of weights.

Definition at line 60 of file APWeightSum.h.

m_k_evt_weight2

double APWeightSum::m_k_evt_weight2

Holds the sum of squared weights.

Definition at line 61 of file APWeightSum.h.

m_k_evt_weight_external

double APWeightSum::m_k_evt_weight_external

Holds the sum of external weights (no trigger weighting).

Definition at line 62 of file APWeightSum.h.

m_linear_uncert

std::vector< THnSparse* > APWeightSum::m_linear_uncert

Holds all histograms for uncertainties.

Definition at line 58 of file APWeightSum.h.

m_variance

double APWeightSum::m_variance

Holds the variance.

Definition at line 63 of file APWeightSum.h.

m_variance_fullcorr

double APWeightSum::m_variance_fullcorr

Holds the variance, assuming full correlation amongst objects.

Definition at line 65 of file APWeightSum.h.

m_variance_nocorr

double APWeightSum::m_variance_nocorr

Holds the variance, assuming no correlations.

Definition at line 64 of file APWeightSum.h.

m_variance_sys

double APWeightSum::m_variance_sys

Holds the systematic variance (from systematics assigned to weights).

Definition at line 66 of file APWeightSum.h.

---

The documentation for this class was generated from the following files:

• APWeightSum.h
• APWeightSum.cxx