Likelihood class that sums over multiple -log components.

The likelihood is often a product of components, for instance when fitting simultaneous pdfs, but also when using subsidiary pdfs. Hence, the negative log likelihood that we, in fact, calculate is often a sum over these components. This sum is implemented by this class.

Definition at line 25 of file RooSumL.h.

Public Member Functions
	RooSumL (RooAbsPdf pdf, RooAbsData data, std::vector< std::unique_ptr< RooAbsL > > components, RooAbsL::Extended extended=RooAbsL::Extended::Auto)

void	constOptimizeTestStatistic (RooAbsArg::ConstOpCode opcode, bool doAlsoTrackingOpt) override
	Interface function signaling a request to perform constant term optimization.

ROOT::Math::KahanSum< double >	evaluatePartition (Section events, std::size_t components_begin, std::size_t components_end) override

std::string	GetClassName () const override

const std::vector< std::unique_ptr< RooAbsL > > &	GetComponents () const

ROOT::Math::KahanSum< double >	getSubsidiaryValue ()

Public Member Functions inherited from RooFit::TestStatistics::RooAbsL
	RooAbsL (ClonePdfData in, std::size_t N_events, std::size_t N_components, Extended extended=Extended::Auto)
	Constructor that clones the pdf/data and owns those cloned copies.

	RooAbsL (const RooAbsL &other)

	RooAbsL (RooAbsPdf pdf, RooAbsData data, std::size_t N_events, std::size_t N_components, Extended extended=Extended::Auto)
	Constructor that does not clone pdf/data and uses the shared_ptr aliasing constructor to make it non-owning.

virtual	~RooAbsL ()=default

virtual double	defaultErrorLevel () const

virtual std::string	GetInfo () const

virtual std::string	GetName () const

std::size_t	getNComponents () const

std::size_t	getNEvents () const

virtual std::unique_ptr< RooArgSet >	getParameters ()

virtual std::string	GetTitle () const

void	initClones (RooAbsPdf &inpdf, RooAbsData &indata)

bool	isExtended () const

virtual std::size_t	numDataEntries () const
	Number of dataset entries.

void	setSimCount (std::size_t value)

Private Attributes
std::vector< std::unique_ptr< RooAbsL > >	components_

Additional Inherited Members
Public Types inherited from RooFit::TestStatistics::RooAbsL
enum class	Extended { Auto , Yes , No }

Static Public Member Functions inherited from RooFit::TestStatistics::RooAbsL
static bool	isExtendedHelper (RooAbsPdf *pdf, Extended extended)

Protected Attributes inherited from RooFit::TestStatistics::RooAbsL
std::shared_ptr< RooAbsData >	data_

bool	extended_ = false

std::size_t	N_components_ = 1

std::size_t	N_events_ = 1

std::unique_ptr< RooArgSet >	normSet_
	Pointer to set with observables used for normalization.

std::shared_ptr< RooAbsPdf >	pdf_

std::size_t	sim_count_ = 1

#include <RooFit/TestStatistics/RooSumL.h>

Inheritance diagram for RooFit::TestStatistics::RooSumL:

[legend]

Constructor & Destructor Documentation

◆ RooSumL()

RooFit::TestStatistics::RooSumL::RooSumL	(	RooAbsPdf *	pdf,
		RooAbsData *	data,
		std::vector< std::unique_ptr< RooAbsL > >	components,
		RooAbsL::Extended	extended = `RooAbsL::Extended::Auto`
	)

Parameters

[in]	pdf	Raw pointer to the pdf; will not be cloned in this object.
[in]	data	Raw pointer to the dataset; will not be cloned in this object.
[in]	components	The component likelihoods.
	extended	Set extended term calculation on, off or use Extended::Auto to determine automatically based on the pdf whether to activate or not.

Warning: components must be passed with std::move, otherwise it cannot be moved into the RooSumL because of the unique_ptr!

Note: The number of events in RooSumL is that of the full dataset. Components will have their own number of events that may be more relevant.

Definition at line 43 of file RooSumL.cxx.

Member Function Documentation

◆ constOptimizeTestStatistic()

void RooFit::TestStatistics::RooSumL::constOptimizeTestStatistic	(	RooAbsArg::ConstOpCode	opcode,
		bool	doAlsoTrackingOpt
	)

overridevirtual

Interface function signaling a request to perform constant term optimization.

The default implementation takes no action other than to forward the calls to all servers. May be overridden in likelihood classes without a cached dataset, like RooSubsidiaryL.

Reimplemented from RooFit::TestStatistics::RooAbsL.

Definition at line 123 of file RooSumL.cxx.

◆ evaluatePartition()

ROOT::Math::KahanSum< double > RooFit::TestStatistics::RooSumL::evaluatePartition	(	Section	events,
		std::size_t	components_begin,
		std::size_t	components_end
	)

overridevirtual

Note: Compared to the RooAbsTestStatistic implementation that this was taken from, we leave out Hybrid and SimComponents interleaving support here. This should be implemented by a calculator (i.e. LikelihoodWrapper or LikelihoodGradientWrapper derived class), if desired.

Implements RooFit::TestStatistics::RooAbsL.

Definition at line 92 of file RooSumL.cxx.

◆ GetClassName()

std::string RooFit::TestStatistics::RooSumL::GetClassName ( ) const

inlineoverridevirtual

Implements RooFit::TestStatistics::RooAbsL.

Definition at line 40 of file RooSumL.h.

◆ GetComponents()

const std::vector< std::unique_ptr< RooAbsL > > & RooFit::TestStatistics::RooSumL::GetComponents ( ) const

inline

Definition at line 42 of file RooSumL.h.

◆ getSubsidiaryValue()

ROOT::Math::KahanSum< double > RooFit::TestStatistics::RooSumL::getSubsidiaryValue ( )

Note: This function assumes there is only one subsidiary component.

Definition at line 112 of file RooSumL.cxx.

Member Data Documentation

◆ components_

std::vector<std::unique_ptr<RooAbsL> > RooFit::TestStatistics::RooSumL::components_

private

Definition at line 45 of file RooSumL.h.

Libraries for RooFit::TestStatistics::RooSumL: