Transformer

Description

Transform a yield onto another observable such as a ratio or a fraction.

Gaussian uncertainties are assumed. Correlations are taken into account and propagated to the new observable using Teddy.

Goal: factorise the filling of the input histogram to the transformation in an observable-agnostic way, hence avoiding multiple executables.

#include <Observable.h>

Inheritance diagram for Transformer:

Public Member Functions
virtual	~Transformer ()=default
virtual void	Transform (const Eigen::VectorXd &) const
virtual void	RemoveBadInputBins (TH1 *, TH2 *)
	Transformer (TUnfoldBinning *, bool=true)

Public Attributes
TUnfoldBinning *	preBinning
TUnfoldBinning *	postBinning

Static Public Attributes
static Eigen::VectorXd	y

Static Protected Member Functions
static void	AddAxis (TUnfoldBinning *, TUnfoldBinning *, int)

Constructor & Destructor Documentation

~Transformer()

virtual ~Transformer ( )

virtualdefault

Destructor.

Transformer()

Transformer	(	TUnfoldBinning *	bng,
		bool	clone_binning = true
	)

Constructor.

Note

subnodes are currently not supported

Parameters

bng	source binning
clone_binning	flag to clone the original binning in target binning

                                                                  :
    preBinning(bng), postBinning(new TUnfoldBinning(bng->GetName()))
{
    postBinning->SetTitle(bng->GetTitle());
 
    if (!clone_binning) return;
 
    int d = bng->GetDistributionDimension();
    for (int i = 0; i < d; ++i)
        AddAxis(preBinning, postBinning, i);
}

Member Function Documentation

AddAxis()

void AddAxis ( TUnfoldBinning *  preBinning, TUnfoldBinning *  postBinning, int  iaxis  )

staticprotected

Copy the axis of the source binning object into the target binning object.

Parameters

preBinning	source binning
postBinning	target binning
iaxis	axis index (follow `TUnfoldBinning`'s convention)

{
    TString axisname = preBinning->GetDistributionAxisLabel(iaxis);
    const TVectorD * binning = preBinning->GetDistributionBinning(iaxis);
    int nbins = binning->GetNrows()-1;
    const double * edges = binning->GetMatrixArray();
    postBinning->AddAxis(axisname, nbins, edges, false, false);
}

RemoveBadInputBins()

void RemoveBadInputBins ( TH1 *  h, TH2 *  cov  )

virtual

Remove bins with too low coverage for Gaussian hypothesis.

Parameters

h	distribution
cov	its covariance matrix

{
    for (int i = preBinning->GetStartBin();
             i < preBinning->GetEndBin(); ++i) {
        double element = h->GetBinContent(i);
 
        // skip empty bins
        if (element == 0) continue;
 
        // avoid two consecutive bins with identical content
        if (h->GetBinContent(i-1) == h->GetBinContent(i)) {
            EmptyBin(h,cov,i);
            continue;
        }
 
        // avoid isolated bins
        if (h->GetBinContent(i-1) == 0 &&
            h->GetBinContent(i+1) == 0)
            EmptyBin(h,cov,i);
    }
}

Transform()

void Transform ( const Eigen::VectorXd &  ) const

virtual

Copies the input distribution by default.

Reimplemented in TestTransformer, and HTnTransformer.

{
    int preOffset =  preBinning->GetStartBin()-1,
       postOffset = postBinning->GetStartBin()-1,
       nBins = preBinning->GetDistributionNumberOfBins();
    y.segment(postOffset, nBins) = x.segment(preOffset, nBins);
}

Member Data Documentation

postBinning

TUnfoldBinning * postBinning

target binning

preBinning

TUnfoldBinning* preBinning

source binning (either rec or gen level)

y

VectorXd y

static

output vector

---

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

• /builds/cms-analysis/general/DasAnalysisSystem/Core/Installer/Core/Unfolding/interface/Observable.h
• /builds/cms-analysis/general/DasAnalysisSystem/Core/Installer/Core/Unfolding/src/Observable.cc