Objective function interface for ROOT Minuit2 minimization. More...

#include <PFitterFcn.h>

Inheritance diagram for PFitterFcn:

Collaboration diagram for PFitterFcn:

[legend]

Public Member Functions
	PFitterFcn (PRunListCollection *runList, Bool_t useChi2)
	Constructor for objective function.

	~PFitterFcn ()
	Destructor.

Double_t	Up () const
	Returns error definition for Minuit2 (Up value).

Double_t	operator() (const std::vector< Double_t > &par) const
	Evaluates objective function for given parameters.

UInt_t	GetTotalNoOfFittedBins ()
	Returns total number of bins used in the fit across all runs.

UInt_t	GetNoOfFittedBins (const UInt_t idx)
	Returns number of fitted bins for a specific run.

void	CalcExpectedChiSquare (const std::vector< Double_t > &par, Double_t &totalExpectedChisq, std::vector< Double_t > &expectedChisqPerRun)
	Calculates expected χ² (or maxLH) for quality assessment.

Private Attributes
Double_t	fUp
	Error definition: 1.0 for χ² (1σ = Δχ²=1), 0.5 for maxLH (1σ = ΔmaxLH=0.5)

Bool_t	fUseChi2
	Fit mode flag: true = χ² minimization, false = max log-likelihood.

PRunListCollection *	fRunListCollection
	Pointer to preprocessed muSR data collection.

Detailed Description

Objective function interface for ROOT Minuit2 minimization.

This class implements the FCNBase interface required by ROOT's Minuit2 minimizer. It provides the objective function (χ² or log-likelihood) that Minuit2 minimizes during parameter optimization.

The class serves as a bridge between musrfit's data structures (PRunListCollection) and Minuit2's optimization algorithms, calculating the goodness-of-fit measure for any given parameter set.

Fitting modes:

χ² minimization: Standard least-squares fitting for Gaussian errors
Maximum likelihood: Poisson statistics, better for low-count data

Usage in fitting workflow:

PFitter creates a PFitterFcn instance with data and fit mode
Minuit2 calls operator()() repeatedly with trial parameter sets
operator()() calculates χ²/maxLH by evaluating theory vs. data
Minuit2 searches parameter space to minimize the returned value
Up() defines the error criterion (Δχ²=1 or ΔmaxLH=0.5 for 1σ)

See also: PFitter, PRunListCollection; ROOT::Minuit2::FCNBase in ROOT Minuit2 documentation

Definition at line 65 of file PFitterFcn.h.

Constructor & Destructor Documentation

◆ PFitterFcn()

PFitterFcn::PFitterFcn	(	PRunListCollection *	runList,
		Bool_t	useChi2 )

Constructor for objective function.

Initializes the function evaluator with preprocessed data and configures the error definition based on the fitting mode.

Parameters

runList	Pointer to collection of preprocessed run data
useChi2	If true, use χ² minimization; if false, use maximum likelihood

Note: The runList pointer must remain valid for the lifetime of this object.

Constructor.

Parameters

runList	run list collection
useChi2	if true, a chisq fit will be performed, otherwise a log max-likelihood fit will be carried out.

Definition at line 41 of file PFitterFcn.cpp.

References fRunListCollection, fUp, and fUseChi2.

◆ ~PFitterFcn()

PFitterFcn::~PFitterFcn ( )

Destructor.

Destructor

Definition at line 59 of file PFitterFcn.cpp.

Member Function Documentation

◆ CalcExpectedChiSquare()

void PFitterFcn::CalcExpectedChiSquare	(	const std::vector< Double_t > &	par,
		Double_t &	totalExpectedChisq,
		std::vector< Double_t > &	expectedChisqPerRun )

Calculates expected χ² (or maxLH) for quality assessment.

Computes the theoretical expected value of χ² assuming the model is correct. This is used to assess goodness-of-fit:

If observed χ² ≈ expected χ²: fit is consistent with data quality
If observed χ² >> expected χ²: systematic deviations present
If observed χ² << expected χ²: possible overestimated errors

For single histogram fits, expected χ² = NDF. For asymmetry fits, the calculation is more complex due to error propagation.

Parameters

par	Parameter vector for evaluation
totalExpectedChisq	Returns total expected χ²/maxLH (output)
expectedChisqPerRun	Returns expected χ²/maxLH for each run (output)

Note: The expectedChisqPerRun vector is resized and filled by this method.

Calculates the expected chisq, expected chisq per run, and chisq per run, if applicable.

Parameters

par
totalExpectedChisq	expected chisq for all run blocks
expectedChisqPerRun	expected chisq vector for all the run blocks

Definition at line 106 of file PFitterFcn.cpp.

References fRunListCollection, and fUseChi2.

◆ GetNoOfFittedBins()

UInt_t PFitterFcn::GetNoOfFittedBins ( const UInt_t idx )

inline

Returns number of fitted bins for a specific run.

Parameters

idx	Run index (0-based)

Returns: Number of bins fitted for the specified run

Definition at line 142 of file PFitterFcn.h.

References fRunListCollection.

◆ GetTotalNoOfFittedBins()

UInt_t PFitterFcn::GetTotalNoOfFittedBins ( )

inline

Returns total number of bins used in the fit across all runs.

Returns: Total count of fitted bins (summed over all runs)

Definition at line 133 of file PFitterFcn.h.

References fRunListCollection.

◆ operator()()

Double_t PFitterFcn::operator() ( const std::vector< Double_t > & par ) const

Evaluates objective function for given parameters.

This is the core function called by Minuit2 during minimization. It computes either χ² or negative log-likelihood by:

Passing parameters to PRunListCollection
Calculating theory predictions for all runs
Comparing theory vs. data across all fitted bins
Returning the total χ²/maxLH value

Parameters

par	Parameter vector with current trial values

Returns: χ² value (if fUseChi2=true) or -2×log-likelihood (if fUseChi2=false)

Note: This function must be const as required by FCNBase interface.; For likelihood fits, returns -2×ln(L) so minimization is equivalent to maximizing L.

Performance:: This function is called hundreds to thousands of times during a fit, so it's optimized for speed (parallel evaluation if OpenMP enabled).

Minuit2 interface function call routine. This is the function which should be minimized.

Parameters

par	a vector with all the parameters of the function

Definition at line 71 of file PFitterFcn.cpp.

References fRunListCollection, and fUseChi2.

◆ Up()

Double_t PFitterFcn::Up ( ) const

inline

Returns error definition for Minuit2 (Up value).

The "Up" value defines what change in the objective function corresponds to 1σ error bars on parameters:

For χ² fits: Up = 1.0 (parabolic errors, Δχ²=1)
For max likelihood: Up = 0.5 (asymmetric errors, ΔmaxLH=0.5)

This value is used by Minuit2's error analysis algorithms (HESSE, MINOS).

Returns: Error definition value (1.0 for χ², 0.5 for likelihood)

See also: ROOT::Minuit2::FCNBase::Up() in Minuit2 manual

Definition at line 102 of file PFitterFcn.h.

References fUp.

Member Data Documentation

◆ fRunListCollection

PRunListCollection* PFitterFcn::fRunListCollection

private

Pointer to preprocessed muSR data collection.

Definition at line 168 of file PFitterFcn.h.

Referenced by CalcExpectedChiSquare(), GetNoOfFittedBins(), GetTotalNoOfFittedBins(), operator()(), and PFitterFcn().

◆ fUp

Double_t PFitterFcn::fUp

private

Error definition: 1.0 for χ² (1σ = Δχ²=1), 0.5 for maxLH (1σ = ΔmaxLH=0.5)

Definition at line 166 of file PFitterFcn.h.

Referenced by PFitterFcn(), and Up().

◆ fUseChi2

Bool_t PFitterFcn::fUseChi2

private

Fit mode flag: true = χ² minimization, false = max log-likelihood.

Definition at line 167 of file PFitterFcn.h.

Referenced by CalcExpectedChiSquare(), operator()(), and PFitterFcn().

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

/workspace/LMU/musrfit/src/include/PFitterFcn.h
/workspace/LMU/musrfit/src/classes/PFitterFcn.cpp

Public Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ PFitterFcn()

◆ ~PFitterFcn()

Member Function Documentation

◆ CalcExpectedChiSquare()

◆ GetNoOfFittedBins()

◆ GetTotalNoOfFittedBins()

◆ operator()()

◆ Up()

Member Data Documentation

◆ fRunListCollection

◆ fUp

◆ fUseChi2