npstat::AbsKDE1DKernel Class Reference

#include <AbsKDE1DKernel.hh>

Inheritance diagram for npstat::AbsKDE1DKernel:

Classes
class	MomentFcn
class	SquareFcn

Public Member Functions
virtual void	setNormFactor (double normfactor)=0
virtual double	xmin () const =0
virtual double	xmax () const =0
virtual double	normFactor () const =0
virtual double	operator() (double x) const =0
virtual AbsKDE1DKernel *	clone () const =0
virtual double	momentIntegral (unsigned k, unsigned nIntegPoints) const
virtual double	squaredIntegral (unsigned nIntegPoints) const
template<typename Numeric >
double	kde (double x, double bandwidth, const Numeric *coords, unsigned long nCoords, bool coordinatesSorted=false) const
template<typename Numeric >
double	reverseKde (double x, double bandwidth, const Numeric *coords, unsigned long nCoords, bool coordinatesSorted=false) const
double	looKde (double bandwidth, unsigned long nCoords, double originalDensityEstimate) const
template<typename Numeric >
double	integratedSquaredError (const AbsDistribution1D &distro, unsigned nIntegIntervals, unsigned nIntegPoints, double bandwidth, bool useReverseKde, const Numeric *coords, unsigned long nCoords, bool coordinatesSorted=false) const
template<typename Numeric >
double	integratedKdeSquared (double xmin, double xmax, unsigned nIntegIntervals, unsigned nIntegPoints, double bandwidth, bool useReverseKde, const Numeric *coords, unsigned long nCoords, bool coordinatesSorted=false) const

Detailed Description

Base class for 1-d kernels and KDE

Member Function Documentation

clone()

virtual AbsKDE1DKernel* npstat::AbsKDE1DKernel::clone ( ) const

pure virtual

"Virtual copy constructor"

Implemented in npstat::KDE1DHOSymbetaKernel, and npstat::KDE1DDensityKernel.

integratedKdeSquared()

template<typename Numeric >

double npstat::AbsKDE1DKernel::integratedKdeSquared	(	double	xmin,
		double	xmax,
		unsigned	nIntegIntervals,
		unsigned	nIntegPoints,
		double	bandwidth,
		bool	useReverseKde,
		const Numeric *	coords,
		unsigned long	nCoords,
		bool	coordinatesSorted = false
	)		const

Integral of KDE squared. Parameter "nIntegPoints" is the number of integration points per interval, and it must be supported by the GaussLegendreQuadrature class.

integratedSquaredError()

template<typename Numeric >

double npstat::AbsKDE1DKernel::integratedSquaredError	(	const AbsDistribution1D &	distro,
		unsigned	nIntegIntervals,
		unsigned	nIntegPoints,
		double	bandwidth,
		bool	useReverseKde,
		const Numeric *	coords,
		unsigned long	nCoords,
		bool	coordinatesSorted = false
	)		const

Integrated squared error w.r.t. a known density. Parameter "nIntegPoints" is the number of integration points per interval, and it must be supported by the GaussLegendreQuadrature class. The integration interval is defined by the support of the density.

kde()

template<typename Numeric >

double npstat::AbsKDE1DKernel::kde	(	double	x,
		double	bandwidth,
		const Numeric *	coords,
		unsigned long	nCoords,
		bool	coordinatesSorted = false
	)		const

Summing K((x - x[i])/h), as in the standard KDE definition. Setting "coordinatesSorted" flag to "true" if coordinates are sorted in the increasing order can potentially increase the speed of KDE evaluation.

looKde()

double npstat::AbsKDE1DKernel::looKde	(	double	bandwidth,
		unsigned long	nCoords,
		double	originalDensityEstimate
	)		const

Leave-one-out density estimate. The original density estimate should be obtained earlier with either "kde" or "reverseKde" method used with the same values of "bandwidth" and "nCoords" arguments.

momentIntegral()

virtual double npstat::AbsKDE1DKernel::momentIntegral ( unsigned  k, unsigned  nIntegPoints  ) const

virtual

"nIntegPoints" argument in the following method is the number of points to use in the quadrature rule. If this method is not overriden, Gauss-Legendre quadrature is used.

reverseKde()

template<typename Numeric >

double npstat::AbsKDE1DKernel::reverseKde	(	double	x,
		double	bandwidth,
		const Numeric *	coords,
		unsigned long	nCoords,
		bool	coordinatesSorted = false
	)		const

Summing K((x[i] - x)/h)

---

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

• stat/AbsKDE1DKernel.hh