Standard psychometric function model. More...

#include <psychometric.h>

Inheritance diagram for PsiPsychometric:

Public Member Functions
	PsiPsychometric (int nAFC, PsiCore core, PsiSigmoid sigmoid)
	Set up a psychometric function model for an nAFC task (nAFC=1 ~> yes/no).
	~PsiPsychometric (void)
	destructor (also deletes the core and sigmoid objects)
virtual double	evaluate (double x, const std::vector< double > &prm) const
	Evaluate the psychometric function at this position.
virtual double	negllikeli (const std::vector< double > &prm, const PsiData *data) const
	negative log likelihood
virtual double	neglpost (const std::vector< double > &prm, const PsiData *data) const
	negative log posterior (unnormalized)
virtual double	leastfavourable (const std::vector< double > &prm, const PsiData *data, double cut, bool threshold=true) const
	derivative of log likelihood in the least favourable direction in parameter space
virtual double	deviance (const std::vector< double > &prm, const PsiData *data) const
	deviance for a given data set and parameter constellation
virtual Matrix *	ddnegllikeli (const std::vector< double > &prm, const PsiData *data) const
	2nd derivative of the negative log likelihood (newly allocated matrix)
virtual std::vector< double >	dnegllikeli (const std::vector< double > &prm, const PsiData *data) const
	1st derivative of the negative log likelihood
const PsiCore *	getCore (void) const
	get the core of the psychometric function
const PsiSigmoid *	getSigmoid (void) const
	get the sigmoid of the psychometric function
void	setPrior (unsigned int index, PsiPrior *prior) throw (BadArgumentError)
	set a Prior for the parameter indicated by index
double	evalPrior (unsigned int index, double x) const
	evaluate the respective prior at value x
virtual double	randPrior (unsigned int index) const
	sample form a prior
const PsiPrior *	getPrior (unsigned int index) const
	get a prior
int	getNalternatives (void) const
	get the number of alternatives (1 means yes/no)
virtual unsigned int	getNparams (void) const
	get the number of free parameters of the psychometric function
virtual std::vector< double >	getStart (const PsiData *data) const
	determine a starting value using logistic regression on a dataset
double	getThres (const std::vector< double > &prm, double cut) const
	get the threshold at a cut between 0 and 1
double	getSlope (const std::vector< double > &prm, double x) const
	get the slope at a stimulus intensity
std::vector< double >	getDevianceResiduals (const std::vector< double > &prm, const PsiData *data) const
	deviance residuals for model checking
double	getRpd (const std::vector< double > &devianceresiduals, const std::vector< double > &prm, const PsiData *data) const
	correlation between deviance residuals and predictions
double	getRkd (const std::vector< double > &devianceresiduals, const PsiData *data) const
	correlation between deviance residuals and block sequence
double	dllikeli (std::vector< double > prm, const PsiData *data, unsigned int i) const
	derivative of the negative loglikelihood with respect to parameter i
double	dlposteri (std::vector< double > prm, const PsiData *data, unsigned int i) const
	derivative of the negative log posterior with respect to parameter i
void	setgammatolambda (void)
double	getGuess (const std::vector< double > &prm) const
	calling this function applies the constraint that gamma and lambda should be equal in a yes/no paradigm
double	dpredict (const std::vector< double > &prm, double x, unsigned int i) const
	partial derivative of psychometric function prediction w.r.t. i-th parameter
double	ddpredict (const std::vector< double > &prm, double x, unsigned int i, unsigned int j) const
	2nd partial derivative of psychometric function prediction w.r.t. i-th and j-th parameters
Protected Member Functions
	PsiPsychometric (int nAFC, PsiCore core, PsiSigmoid sigmoid, unsigned int nparameters)
	Set up a psychometric function model for an nAFC task, explicitely specifiing the number of parameters (useful for derived classes).

Detailed Description

Standard psychometric function model.

Standard model for the psychometric function that assumes that the number of correct responses is a binomial random variable with parameters N (number of trials) and Psi, where Psi is

Psi = guessingrate + (1-guessingrate-lapserate) * Sigmoid ( x | theta )

For an nAFC task, the guessingrate is typicall fixed at 1/n.

The term Sigmoid ( x | theta ) is represented by two objects: a PsiSigmoid f, that describes a nonlinear function from the real numbers to (0,1) and a PsiCore g, that describes the "internal" workings of the nonlinear function. Thus, the Term Sigmoid ( x | theta ) equals f(g(x,theta)).

The object copies and manages its own memory. This means that when setting PsiSigmoid and PsiCore in the constructor, as well as setting PsiPrior via setPrior(), copies of the respective objects are made. Furthermore when the object is destroyed theses copies are destoryed also.

Constructor & Destructor Documentation

PsiPsychometric::PsiPsychometric	(	int	nAFC,
		PsiCore *	core,
		PsiSigmoid *	sigmoid,
		unsigned int	nparameters
	)			`[protected]`

Set up a psychometric function model for an nAFC task, explicitely specifiing the number of parameters (useful for derived classes).

Parameters:

	nAFC	number of alternatives (1 indicating yes/no)
	core	internal part of the nonlinear function
	sigmoid	"external" saturating part of the nonlinear function
	nparameters	number of parameters given explicitely

PsiPsychometric::PsiPsychometric	(	int	nAFC,
		PsiCore *	core,
		PsiSigmoid *	sigmoid
	)

Set up a psychometric function model for an nAFC task (nAFC=1 ~> yes/no).

Parameters:

	nAFC	number of alternatives in the task (1 indicating yes/no)
	core	internal part of the nonlinear function (in many cases this is actually a linear function)
	sigmoid	"external" saturating part of the nonlinear function

Member Function Documentation

Matrix * PsiPsychometric::ddnegllikeli	(	const std::vector< double > &	prm,
		const PsiData *	data
	)			const `[virtual]`

2nd derivative of the negative log likelihood (newly allocated matrix)

Parameters:

	prm	parameters at which the second derivative should be evaluated
	data	data for which the likelihood should be evaluated

Reimplemented in BetaPsychometric.

double PsiPsychometric::deviance	(	const std::vector< double > &	prm,
		const PsiData *	data
	)			const `[virtual]`

deviance for a given data set and parameter constellation

Parameters:

	prm	parameters of the psychometric functin model
	data	data for which the likelihood should be evaluated

Reimplemented in BetaPsychometric, and OutlierModel.

double PsiPsychometric::dllikeli	(	std::vector< double >	prm,
		const PsiData *	data,
		unsigned int	i
	)			const

derivative of the negative loglikelihood with respect to parameter i

Parameters:

	prm	parameters of the model
	data	data for which the likelihood should be evaluated
	i	index of the parameter for which the derivative should be evaluated

double PsiPsychometric::dlposteri	(	std::vector< double >	prm,
		const PsiData *	data,
		unsigned int	i
	)			const

derivative of the negative log posterior with respect to parameter i

Parameters:

	prm	parameters of the psychometric function model
	data	data for which the likelihood should be valuated
	i	index of the parameter for which the derivative should be evaluated

std::vector< double > PsiPsychometric::dnegllikeli	(	const std::vector< double > &	prm,
		const PsiData *	data
	)			const `[virtual]`

1st derivative of the negative log likelihood

Parameters:

	prm	parameters at which the first derivative should be evaluated
	data	data for which the likelihood should be evaluated

Reimplemented in BetaPsychometric.

double PsiPsychometric::evaluate	(	double	x,
		const std::vector< double > &	prm
	)			const `[virtual]`

Evaluate the psychometric function at this position.

Parameters:

	x	stimulus intensity
	prm	parameters of the psychometric function model

std::vector< double > PsiPsychometric::getDevianceResiduals	(	const std::vector< double > &	prm,
		const PsiData *	data
	)			const

deviance residuals for model checking

Parameters:

	prm	parameters of the psychometric function model
	data	data for which the deviance residuals should be determined

double PsiPsychometric::getRpd	(	const std::vector< double > &	devianceresiduals,
		const std::vector< double > &	prm,
		const PsiData *	data
	)			const

correlation between deviance residuals and predictions

Parameters:

	devianceresiduals	deviance residuals as determined by getDevianceResiduals
	prm	parameters of the psychometric function model
	data	data set corresponding to the deviance residuals

double PsiPsychometric::getSlope	(	const std::vector< double > &	prm,
		double	x
	)			const `[inline]`

get the slope at a stimulus intensity

Parameters:

	prm	parameters of the psychometric function model
	x	performance level at which the slope should be evaluated

double PsiPsychometric::getThres	(	const std::vector< double > &	prm,
		double	cut
	)			const `[inline]`

get the threshold at a cut between 0 and 1

Parameters:

	prm	parameters of the psychometric function model
	cut	performance level at which the threshold should be evaluated

double PsiPsychometric::leastfavourable	(	const std::vector< double > &	prm,
		const PsiData *	data,
		double	cut,
		bool	threshold = `true`
	)			const `[virtual]`

derivative of log likelihood in the least favourable direction in parameter space

Parameters:

	prm	parameters of the psychometric function model
	data	data for which the likelihood should be evaluated
	cut	performance level at which the threshold should be evaluated
	threshold	should the calculations be performed for thresholds? (anything else is not yet implemented)

double PsiPsychometric::negllikeli	(	const std::vector< double > &	prm,
		const PsiData *	data
	)			const `[virtual]`

negative log likelihood

Parameters:

	prm	parameters of the psychometric function model
	data	data for which the likelihood should be evaluated

Reimplemented in BetaPsychometric, and OutlierModel.

double PsiPsychometric::neglpost	(	const std::vector< double > &	prm,
		const PsiData *	data
	)			const `[virtual]`

negative log posterior (unnormalized)

Parameters:

	prm	parameters of the psychometric function model
	data	data for which the posterior should be evaluated

Reimplemented in OutlierModel.

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

src/psychometric.h
src/psychometric.cc

PsiPsychometric Class Reference

Public Member Functions

Protected Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation