PEX inverse-covariance. More...

#include <MuGen.h>

Inheritance diagram for SigmaIpex:

Collaboration diagram for SigmaIpex:

Public Member Functions
	SigmaIpex ()
	Default constructor.

	SigmaIpex (const size_t &d, const double &invVar, const double &df)
	Deterministic diagonal matrix constructor. More...

	SigmaIpex (const size_t &d, const double &invVar, const double &df, const string &outFlNam)
	Deterministic diagonal matrix constructor with output file name. More...

	SigmaIpex (const Grp &dat, const double &prDiag, const double &nu0)
	Location data-based constructor. More...

	SigmaIpex (const Grp &dat, const string &outFlNam, const double &prDiag, const double &nu0)
	Location data-based constructor with output file name. More...

	~SigmaIpex ()
	Destructor.

void	save (const char *how="a")
	Save to a stored file name. More...

void	save (const string &fileNam, const char *how="a")
	Save to a given file name. More...

void	update (const Grp &dat, const Qgrp &q)
	Basic Student- \(t\) update. More...

void	update (const Grp &dat, const Grp &mu, const Qgrp &q)
	Student- \(t\) update with a mean. More...

Public Member Functions inherited from SigmaI
	SigmaI ()
	Default constructor. More...

	SigmaI (const size_t &d, const double &invVar)
	Deterministic diagonal matrix constructor. More...

	SigmaI (const size_t &d, const double &invVar, const double &df)
	Deterministic diagonal matrix constructor. More...

	SigmaI (const size_t &d, const double &invVar, const double &df, const string &outFlNam)
	Deterministic diagonal matrix constructor with output file name. More...

	SigmaI (const gsl_matrix *mat)
	Deterministic matrix constructor. More...

	SigmaI (const gsl_matrix S, const size_t &d, const size_t &df, const gsl_matrix LamPr, const double &nu0)
	Constructor with a matrix prior. More...

	SigmaI (const gsl_matrix *S, const size_t &d, const size_t &df, const double &diagPr, const double &nu0)
	Constructor with a diagonal prior. More...

	SigmaI (const Grp &dat, const double &prDiag, const double &nu0)
	Location data-based constructor. More...

	SigmaI (const Grp &dat, const string &outFlNam, const double &prDiag, const double &nu0)
	Location data-based constructor with output file name. More...

	SigmaI (const SigmaI &)
	Copy constructor. More...

SigmaI &	operator= (const SigmaI &)
	Assignment operator. More...

virtual	~SigmaI ()
	Destructor.

virtual void	update (const Grp &dat)
	Basic Gaussian update. More...

virtual void	update (const Grp &dat, const Grp &mu)
	Gaussian update with a mean. More...

void	save (const string &fileNam, const Apex &A, const char *how="a")
	Save adjusted matrix. More...

void	save (FILE *fileStr)
	Save to file stream. More...

string	getOutFile () const
	Access the output file. More...

const gsl_matrix *	getMat () const
	Access to the inverse-covariance matrix. More...

void	srDetUpdate ()
	Update square-root of the determinant.

double	getSrDet () const
	Access the square-root of the determinant. More...

Protected Attributes
double	_alpha
	The PEX \( \alpha \) parameter.

Protected Attributes inherited from SigmaI
gsl_matrix *	_mat
	Inverse-covariance matrix. More...

size_t	_d
	Dimension of the matrix. More...

double	_srDet
	Square root of the determinant. More...

gsl_matrix *	_LamSc
	Prior inverse-covariance. More...

double	_n0
	Prior degrees of freedom.

string	_outFlNam
	Output file name.

gsl_rng *	_r
	Pseudo-random number generator. More...

Detailed Description

PEX inverse-covariance.

Implementation of the van Dyk and Meng's [dyk01] parameter expansion scheme from multivariate Student- \(t\) sampling. Note that it only works when all the scale values in Qgrp are sampled. So, for example, when we have missing data with a Student- \(t\) model for the rest (see Qgrp documentation for details), this class should not be used, it will cause divergence and crashing.

Note: Not to be confused with location parameter PEX such as MuGrpPEX

Constructor & Destructor Documentation

◆ SigmaIpex() [1/4]

SigmaIpex::SigmaIpex	(	const size_t &	d,
		const double &	invVar,
		const double &	df
	)

inline

Deterministic diagonal matrix constructor.

Creates a diagonal matrix with the provided inverse-variance and degrees of freedom parameter.

Parameters

[in]	size_t&	number of rows and columns
[in]	double&	inverse variance
[in]	double&	degrees of freedom

◆ SigmaIpex() [2/4]

SigmaIpex::SigmaIpex	(	const size_t &	d,
		const double &	invVar,
		const double &	df,
		const string &	outFlNam
	)

inline

Deterministic diagonal matrix constructor with output file name.

Creates a diagonal matrix with the provided inverse-variance and degrees of freedom parameter.

Parameters

[in]	size_t&	number of rows and columns
[in]	double&	inverse variance
[in]	double&	degrees of freedom
[in]	string&	output file name

◆ SigmaIpex() [3/4]

SigmaIpex::SigmaIpex	(	const Grp &	dat,
		const double &	prDiag,
		const double &	nu0
	)

inline

Location data-based constructor.

Initializes the object with an inverse-covariance of the provided location data. The dimension of the object os equal to the number of columns in the data.

Parameters

[in]	Grp&	data
[in]	double&	prior inverse-variance
[in]	double&	prior degrees of freedom

◆ SigmaIpex() [4/4]

SigmaIpex::SigmaIpex	(	const Grp &	dat,
		const string &	outFlNam,
		const double &	prDiag,
		const double &	nu0
	)

inline

Location data-based constructor with output file name.

Initializes the object with an inverse-covariance of the provided location data. The dimension of the object os equal to the number of columns in the data.

Parameters

[in]	Grp&	data
[in]	string&	output file name
[in]	double&	prior inverse-variance
[in]	double&	prior degrees of freedom

Member Function Documentation

◆ save() [1/2]

void SigmaIpex::save ( const char * how = "a" )

virtual

Save to a stored file name.

Saves the "adjusted" value, i.e. scaled by the multiplicative parameter.

Parameters

[in] char* saving mode, appending by default

Reimplemented from SigmaI.

◆ save() [2/2]

void SigmaIpex::save	(	const string &	fileNam,
		const char *	how = `"a"`
	)

virtual

Save to a given file name.

Saves the "adjusted" value, i.e. scaled by the multiplicative parameter.

Parameters

[in]	string&	output file name
[in]	char*	saving mode, appending by default

Reimplemented from SigmaI.

◆ update() [1/2]

void SigmaIpex::update	(	const Grp &	dat,
		const Grp &	mu,
		const Qgrp &	q
	)

virtual

Student- \(t\) update with a mean.

The mean value is subtracted from the data according to the up-pointing RanIndex in the data object.

Parameters

[in]	Grp&	data
[in]	Grp&	mean
[in]	Qgrp&	scale parameter

Reimplemented from SigmaI.

◆ update() [2/2]

void SigmaIpex::update	(	const Grp &	dat,
		const Qgrp &	q
	)

virtual

Basic Student- \(t\) update.

The data are assumed already centered, so the update is based on the simple cross-product of the data and the current value of the scale parameter.

Parameters

[in]	Grp&	data
[in]	Qgrp&	scale parameter

Reimplemented from SigmaI.

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

Public Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ SigmaIpex() [1/4]

◆ SigmaIpex() [2/4]

◆ SigmaIpex() [3/4]

◆ SigmaIpex() [4/4]

Member Function Documentation

◆ save() [1/2]

◆ save() [2/2]

◆ update() [1/2]

◆ update() [2/2]