Coordinates statistical inputs, nuisance distributions, MLE fits and contour/scan computations. More...

#include <StatisticManager.h>

Public Member Functions
	StatisticManager (StatisticConfig config, std::shared_ptr< IModel > obs_int, std::shared_ptr< IStatCorrelationProxy > pscp, std::shared_ptr< IStatParameterProxy > pspp, std::shared_ptr< IStatSourcesProxy > sp, std::shared_ptr< IStatDependencyPruner > dp, std::shared_ptr< INuisanceReader > nuisance_reader, std::shared_ptr< IStatParamOptimizerProxy > spop)
	Constructs a statistic manager and initializes observable/nuisance state.

std::vector< std::unique_ptr< IMarginalDistribution > >	build_nuisance_marginal_distributions ()
	Builds marginal distributions for all currently cached nuisances.

std::unique_ptr< JointDistribution >	build_nuisance_distribution ()
	Builds the joint nuisance distribution from cached nuisance marginals and correlations.

std::unique_ptr< JointDistribution >	build_exp_data_distribution ()
	Builds the joint experimental-data distribution from cached observables and correlations.

void	reload_nuisance_specs ()
	Reloads default and user nuisance specifications and invalidates fit state.

void	set_nuisance_user_file (const fs::path &user_yaml_path)
	Selects a custom user nuisance-definition file and reloads nuisance specifications.

void	clear_nuisance_user_file ()
	Clears the custom user nuisance file and reloads the default configured user file.

const NuisanceRegistry &	default_nuisance_specs () const

const NuisanceRegistry &	user_nuisance_specs () const

const NuisanceRegistry &	merged_nuisance_specs () const

std::map< BinnedObservableId, GaussianSummary >	compute_uncertainties ()
	Computes Gaussian summaries for MC-propagated observable uncertainties.

MCResult	compute_uncertainties_and_sampling ()
	Runs MC uncertainty propagation and returns both samples and summaries.

FitResultWithMaps	compute_MLE (const std::vector< ParamId > &p_specs)
	Computes the maximum-likelihood fit for a selected set of fit parameters.

Contour	confidence_contour (ParamId p1, ParamId p2, double z, std::array< double, 4 > bounds, ContourOptions options)
	Computes a two-dimensional confidence contour for the last successful MLE.

void	prepare_likelihood_for_scan (const std::vector< ParamId > &p_specs)
	Prepares a likelihood object for manual scans without running a full MLE.

void	set_manual_scan_point (const std::map< ParamId, double > &p_hat, const std::map< ParamId, double > &eta_hat)
	Sets the reference point used by subsequent likelihood scans.

LikelihoodScanGrid	scan_likelihood_around_current_point (ParamId p1, ParamId p2, double x_half_width, double y_half_width, std::size_t nx, std::size_t ny) const
	Evaluates the current likelihood on a regular 2D grid around the active reference point.

void	save_likelihood_scan_csv (const std::string &path, const LikelihoodScanGrid &grid) const
	Writes a likelihood scan grid to a CSV file.

void	update_cache (const std::vector< ParamId > &p_specs=std::vector< ParamId >())
	Updates the full statistical cache for the selected fit parameters.

void	select_experiment (const std::string &experiment)
	Restricts subsequent statistics to a single experiment name.

void	select_experiments (const std::set< std::string > &experiments)
	Restricts subsequent statistics to the provided set of experiment names.

void	select_experiments (const std::vector< std::string > &experiments)
	Restricts subsequent statistics to the provided vector of experiment names.

void	select_experiments_all ()
	Clears any experiment selection and uses all available experiments.

bool	has_experiment_selection () const noexcept

std::set< std::string >	selected_experiments () const

void	select_experiment_observables (const std::set< ExperimentObs > &observables)
	Restricts subsequent statistics to an explicit list of experimental measurements.

void	select_experiment_observables (const std::vector< ExperimentObs > &observables)
	Vector overload for select_experiment_observables.

void	select_experiment_observables_all ()
	Clears the explicit experimental-observable selection.

bool	has_experiment_observable_selection () const noexcept

std::set< ExperimentObs >	selected_experiment_observables () const

std::map< ParamId, double >	get_all_obss_deps ()
	Selects all nuisance dependencies relevant to the current observable set.

std::map< ParamId, double >	get_p_specs (const std::vector< ParamId > &p_specs)
	Resolves initial fit-parameter values from the parameter proxy.

std::map< ParamId, std::map< ParamId, double > >	get_all_correlations ()

std::map< ExperimentObs, std::map< ExperimentObs, double > >	get_all_obs_correlations ()

std::map< ExperimentObs, double >	get_obs_exp ()

void	print_cache ()
	Prints the current internal cache to standard output for debugging.

Detailed Description

Coordinates statistical inputs, nuisance distributions, MLE fits and contour/scan computations.

The manager owns no low-level physics data itself. Instead, it delegates model predictions to IModel, parameter values and uncertainties to the statistical proxies, and nuisance metadata to INuisanceReader. Its role is to assemble those sources into the distributions and likelihoods required by the fitting workflow.

Definition at line 210 of file StatisticManager.h.

Constructor & Destructor Documentation

◆ StatisticManager()

StatisticManager::StatisticManager	(	StatisticConfig	config,
		std::shared_ptr< IModel >	obs_int,
		std::shared_ptr< IStatCorrelationProxy >	pscp,
		std::shared_ptr< IStatParameterProxy >	pspp,
		std::shared_ptr< IStatSourcesProxy >	sp,
		std::shared_ptr< IStatDependencyPruner >	dp,
		std::shared_ptr< INuisanceReader >	nuisance_reader,
		std::shared_ptr< IStatParamOptimizerProxy >	spop
	)

Constructs a statistic manager and initializes observable/nuisance state.

The constructor computes the model observables once, loads default and user nuisance specifications, merges them, and invalidates any previous fit state.

Parameters

config	Runtime statistical configuration.
obs_int	Model used to compute and predict observables.
pscp	Correlation proxy for parameters and observables.
pspp	Parameter proxy providing central values and uncertainties.
sp	Source proxy used to resolve leaf dependencies.
dp	Dependency pruner used to detach fit parameters from nuisance dependencies.
nuisance_reader	Reader used to load default/user nuisance specifications.
spop	Optional/statistical parameter optimizer proxy used by the surrounding workflow.

Exceptions

std::invalid_argument if nuisance_reader is null.

Definition at line 514 of file StatisticManager.cpp.

Member Function Documentation

◆ build_exp_data_distribution()

std::unique_ptr< JointDistribution > StatisticManager::build_exp_data_distribution ( )

Builds the joint experimental-data distribution from cached observables and correlations.

Returns: Joint distribution over selected experimental observables.

Definition at line 601 of file StatisticManager.cpp.

◆ build_nuisance_distribution()

std::unique_ptr< JointDistribution > StatisticManager::build_nuisance_distribution ( )

Builds the joint nuisance distribution from cached nuisance marginals and correlations.

Returns: Joint distribution over selected nuisance parameters.

Definition at line 573 of file StatisticManager.cpp.

◆ build_nuisance_marginal_distributions()

std::vector< std::unique_ptr< IMarginalDistribution > > StatisticManager::build_nuisance_marginal_distributions ( )

Builds marginal distributions for all currently cached nuisances.

Returns: Marginal distributions ordered consistently with StatCache::eta_specs_real.

Definition at line 541 of file StatisticManager.cpp.

◆ clear_nuisance_user_file()

void StatisticManager::clear_nuisance_user_file ( )

Clears the custom user nuisance file and reloads the default configured user file.

Definition at line 1284 of file StatisticManager.cpp.

◆ compute_MLE()

FitResultWithMaps StatisticManager::compute_MLE ( const std::vector< ParamId > & p_specs )

Computes the maximum-likelihood fit for a selected set of fit parameters.

The likelihood backend is selected through StatisticConfig::likelihood_mode. This method also updates the internal state required by contour and scan methods.

Parameters

p_specs Ordered list of fit-parameter identifiers.

Returns: Map-based fit result.

Exceptions

std::invalid_argument if p_specs resolves to an empty fit-parameter set.

Definition at line 683 of file StatisticManager.cpp.

◆ compute_uncertainties()

std::map< BinnedObservableId, GaussianSummary > StatisticManager::compute_uncertainties ( )

Computes Gaussian summaries for MC-propagated observable uncertainties.

Returns: Map from binned observable identifiers to Gaussian summaries.

Definition at line 642 of file StatisticManager.cpp.

◆ compute_uncertainties_and_sampling()

MCResult StatisticManager::compute_uncertainties_and_sampling ( )

Runs MC uncertainty propagation and returns both samples and summaries.

Returns: Full MC result including accepted samples, Gaussian summaries and covariance.

Definition at line 668 of file StatisticManager.cpp.

◆ confidence_contour()

Contour StatisticManager::confidence_contour	(	ParamId	p1,
		ParamId	p2,
		double	z,
		std::array< double, 4 >	bounds,
		ContourOptions	options
	)

Computes a two-dimensional confidence contour for the last successful MLE.

Parameters

p1	First fit parameter.
p2	Second fit parameter.
z	Gaussian-equivalent significance level.
bounds	Bounds as {xmin, xmax, ymin, ymax}.
options	Contour and profiling options.

Returns: Extracted contour.

Exceptions

std::runtime_error	if no MLE has been computed.
std::invalid_argument	if the parameters are not part of the last fit or are identical.

Definition at line 949 of file StatisticManager.cpp.

◆ default_nuisance_specs()

const NuisanceRegistry & StatisticManager::default_nuisance_specs ( ) const

inline

Returns: Default nuisance specifications loaded from the default nuisance file.

Definition at line 273 of file StatisticManager.h.

◆ get_all_correlations()

std::map< ParamId, std::map< ParamId, double > > StatisticManager::get_all_correlations ( )

Returns: Correlation matrix for the currently cached nuisance parameters.

Definition at line 1700 of file StatisticManager.cpp.

◆ get_all_obs_correlations()

std::map< ExperimentObs, std::map< ExperimentObs, double > > StatisticManager::get_all_obs_correlations ( )

Returns: Correlation matrix for the currently cached experimental observables.

Definition at line 1707 of file StatisticManager.cpp.

◆ get_all_obss_deps()

std::map< ParamId, double > StatisticManager::get_all_obss_deps ( )

Selects all nuisance dependencies relevant to the current observable set.

Returns: Map of selected nuisance identifiers to central values.

Definition at line 1382 of file StatisticManager.cpp.

◆ get_obs_exp()

std::map< ExperimentObs, double > StatisticManager::get_obs_exp ( )

Returns: Experimental central values selected for the current observable/experiment set.

Definition at line 1715 of file StatisticManager.cpp.

◆ get_p_specs()

std::map< ParamId, double > StatisticManager::get_p_specs ( const std::vector< ParamId > & p_specs )

Resolves initial fit-parameter values from the parameter proxy.

Parameters

p_specs Fit-parameter identifiers.

Returns: Map of parameter identifiers to central values.

Definition at line 1693 of file StatisticManager.cpp.

◆ has_experiment_observable_selection()

bool StatisticManager::has_experiment_observable_selection ( ) const

noexcept

Returns: True if an explicit experimental-observable selection is active.

Definition at line 2045 of file StatisticManager.cpp.

◆ has_experiment_selection()

bool StatisticManager::has_experiment_selection ( ) const

noexcept

Returns: True if an experiment selection is currently active.

Definition at line 1996 of file StatisticManager.cpp.

◆ merged_nuisance_specs()

const NuisanceRegistry & StatisticManager::merged_nuisance_specs ( ) const

inline

Returns: Merged nuisance specifications, with user entries overriding default ones.

Definition at line 279 of file StatisticManager.h.

◆ prepare_likelihood_for_scan()

void StatisticManager::prepare_likelihood_for_scan ( const std::vector< ParamId > & p_specs )

Prepares a likelihood object for manual scans without running a full MLE.

The current central values of the fit parameters and nuisance parameters are stored as the default scan reference point. A later call to compute_MLE(...) or set_manual_scan_point(...) overrides this reference.

Parameters

p_specs Ordered list of fit parameters to expose in the scan.

Definition at line 1733 of file StatisticManager.cpp.

◆ print_cache()

void StatisticManager::print_cache ( )

Prints the current internal cache to standard output for debugging.

Definition at line 1172 of file StatisticManager.cpp.

◆ reload_nuisance_specs()

void StatisticManager::reload_nuisance_specs ( )

Reloads default and user nuisance specifications and invalidates fit state.

Definition at line 1266 of file StatisticManager.cpp.

◆ save_likelihood_scan_csv()

void StatisticManager::save_likelihood_scan_csv	(	const std::string &	path,
		const LikelihoodScanGrid &	grid
	)		const

Writes a likelihood scan grid to a CSV file.

Parameters

path	Output CSV path.
grid	Scan grid to serialize.

Definition at line 1949 of file StatisticManager.cpp.

◆ scan_likelihood_around_current_point()

LikelihoodScanGrid StatisticManager::scan_likelihood_around_current_point	(	ParamId	p1,
		ParamId	p2,
		double	x_half_width,
		double	y_half_width,
		std::size_t	nx,
		std::size_t	ny
	)		const

Evaluates the current likelihood on a regular 2D grid around the active reference point.

Parameters

p1	First scanned parameter.
p2	Second scanned parameter.
x_half_width	Half-width of the scan window along `p1`.
y_half_width	Half-width of the scan window along `p2`.
nx	Number of x grid points.
ny	Number of y grid points.

Returns: Filled scan grid with NLL and delta-NLL values.

Definition at line 1841 of file StatisticManager.cpp.

◆ select_experiment()

void StatisticManager::select_experiment ( const std::string & experiment )

Restricts subsequent statistics to a single experiment name.

Definition at line 1969 of file StatisticManager.cpp.

◆ select_experiment_observables() [1/2]

void StatisticManager::select_experiment_observables ( const std::set< ExperimentObs > & observables )

Restricts subsequent statistics to an explicit list of experimental measurements.

This is stricter than select_experiments(): the experiment-name filter keeps every measurement from the selected experiments whose theory observable has been registered, whereas this method keeps only the exact (experiment, observable/bin) entries listed here. It is useful to reproduce SuperIso myobs.in / arXiv ancillary observable lists exactly.

Definition at line 2029 of file StatisticManager.cpp.

◆ select_experiment_observables() [2/2]

void StatisticManager::select_experiment_observables ( const std::vector< ExperimentObs > & observables )

Vector overload for select_experiment_observables.

Definition at line 2022 of file StatisticManager.cpp.

◆ select_experiment_observables_all()

void StatisticManager::select_experiment_observables_all ( )

Clears the explicit experimental-observable selection.

Definition at line 2040 of file StatisticManager.cpp.

◆ select_experiments() [1/2]

void StatisticManager::select_experiments ( const std::set< std::string > & experiments )

Restricts subsequent statistics to the provided set of experiment names.

Definition at line 1980 of file StatisticManager.cpp.

◆ select_experiments() [2/2]

void StatisticManager::select_experiments ( const std::vector< std::string > & experiments )

Restricts subsequent statistics to the provided vector of experiment names.

Definition at line 1973 of file StatisticManager.cpp.

◆ select_experiments_all()

void StatisticManager::select_experiments_all ( )

Clears any experiment selection and uses all available experiments.

Definition at line 1991 of file StatisticManager.cpp.

◆ selected_experiment_observables()

std::set< ExperimentObs > StatisticManager::selected_experiment_observables ( ) const

Returns: Active explicit experimental-observable selection, or an empty set when inactive.

Definition at line 2049 of file StatisticManager.cpp.

◆ selected_experiments()

std::set< std::string > StatisticManager::selected_experiments ( ) const

Returns: Active experiment selection, or an empty set when all experiments are selected.

Definition at line 2000 of file StatisticManager.cpp.

◆ set_manual_scan_point()

void StatisticManager::set_manual_scan_point	(	const std::map< ParamId, double > &	p_hat,
		const std::map< ParamId, double > &	eta_hat
	)

Sets the reference point used by subsequent likelihood scans.

Parameters

p_hat	Map of fit-parameter values at the manual reference point.
eta_hat	Map of nuisance values at the manual reference point.

Exceptions

std::runtime_error	if no likelihood has been prepared.
std::invalid_argument	if one required parameter value is missing.

Definition at line 1809 of file StatisticManager.cpp.

◆ set_nuisance_user_file()

void StatisticManager::set_nuisance_user_file ( const fs::path & user_yaml_path )

Selects a custom user nuisance-definition file and reloads nuisance specifications.

Parameters

user_yaml_path Path to the user YAML/JSON nuisance file.

Definition at line 1279 of file StatisticManager.cpp.

◆ update_cache()

void StatisticManager::update_cache ( const std::vector< ParamId > & p_specs = std::vector<ParamId>() )

Updates the full statistical cache for the selected fit parameters.

Parameters

p_specs Fit parameters to detach from the nuisance set.

Definition at line 1181 of file StatisticManager.cpp.

◆ user_nuisance_specs()

const NuisanceRegistry & StatisticManager::user_nuisance_specs ( ) const

inline

Returns: User nuisance specifications loaded from the active user nuisance file.

Definition at line 276 of file StatisticManager.h.

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

/home/runner/work/Hyperiso/Hyperiso/Hyperiso/Hyperiso/core/src/Statistic/domain/StatisticManager.h
/home/runner/work/Hyperiso/Hyperiso/Hyperiso/Hyperiso/core/src/Statistic/domain/StatisticManager.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ StatisticManager()

Member Function Documentation

◆ build_exp_data_distribution()

◆ build_nuisance_distribution()

◆ build_nuisance_marginal_distributions()

◆ clear_nuisance_user_file()

◆ compute_MLE()

◆ compute_uncertainties()

◆ compute_uncertainties_and_sampling()

◆ confidence_contour()

◆ default_nuisance_specs()

◆ get_all_correlations()

◆ get_all_obs_correlations()

◆ get_all_obss_deps()

◆ get_obs_exp()

◆ get_p_specs()

◆ has_experiment_observable_selection()

◆ has_experiment_selection()

◆ merged_nuisance_specs()

◆ prepare_likelihood_for_scan()

◆ print_cache()

◆ reload_nuisance_specs()

◆ save_likelihood_scan_csv()

◆ scan_likelihood_around_current_point()

◆ select_experiment()

◆ select_experiment_observables() [1/2]

◆ select_experiment_observables() [2/2]

◆ select_experiment_observables_all()

◆ select_experiments() [1/2]

◆ select_experiments() [2/2]

◆ select_experiments_all()

◆ selected_experiment_observables()

◆ selected_experiments()

◆ set_manual_scan_point()

◆ set_nuisance_user_file()

◆ update_cache()

◆ user_nuisance_specs()