npstat::AbsCVCopulaSmoother Class Reference

Inheritance diagram for npstat::AbsCVCopulaSmoother:

Public Member Functions
bool	isConvolving () const
void	setConvolving (const bool b)
const std::vector< double > &	bandwidthValues () const
const std::vector< double > &	lastCVValues () const
const std::vector< double > &	lastRegularizedFractions () const
unsigned	getNFilters () const
unsigned	lastFilterChosen () const
Public Member Functions inherited from npstat::AbsCopulaSmootherBase
	AbsCopulaSmootherBase (const unsigned *nBinsInEachDim, unsigned dim, double tolerance, unsigned maxNormCycles)
unsigned	dim () const
ArrayShape	copulaShape () const
void	setArchive (gs::AbsArchive *ar, const char *category=0)
template<class Point >
const HistoND< double > &	smooth (unsigned long uniqueId, std::vector< OrderedPointND< Point > > &in, double *bandwidthUsed=0)
template<class Point >
const HistoND< double > &	weightedSmooth (unsigned long uniqueId, const std::vector< std::pair< const Point *, double > > &in, const unsigned *dimsToUse, unsigned nDimsToUse, double *bandwidthUsed=0)

Protected Member Functions
	AbsCVCopulaSmoother (const unsigned *nBinsInEachDim, unsigned dim, double marginTolerance, unsigned maxNormCycles, double initialBw, double cvRange, unsigned nCV, bool useConvolve)
	AbsCVCopulaSmoother (const unsigned *nBinsInEachDim, unsigned dim, double marginTolerance, unsigned maxNormCycles, const std::vector< double > &bandwidthValues, bool useConvolve)

Constructor & Destructor Documentation

AbsCVCopulaSmoother() [1/2]

npstat::AbsCVCopulaSmoother::AbsCVCopulaSmoother ( const unsigned *  nBinsInEachDim, unsigned  dim, double  marginTolerance, unsigned  maxNormCycles, double  initialBw, double  cvRange, unsigned  nCV, bool  useConvolve  )

protected

Constructor arguments are as follows:

nBinsInEachDim – number of copula bins in each dimension

dim – copula dimensionality

marginTolerance – tolerance for the margin to be uniform

maxNormCycles – max number of copula normalization cycles

initialBw – "central" bandwidth for cross validation calculations (or the actual bandwidth used in case cross validation is not performed). Set this parameter to 0.0 in order to disable filtering altogether.

cvRange – we will scan bandwidth values between initialBw/cvRange and initialBw*cvRange uniformly in the log space.

nCV – number of bandwidth values to try in the bandwidth scan. If this number is even, it will be increased by 1 internally so that the "central" bandwidth is included in the scan. If this parameter is 0 or 1, the value given by "initialBw" will be used and cross-validation will not be performed.

useConvolve – if "true", use "convolve" method of the filter rather than "filter" method.

AbsCVCopulaSmoother() [2/2]

npstat::AbsCVCopulaSmoother::AbsCVCopulaSmoother ( const unsigned *  nBinsInEachDim, unsigned  dim, double  marginTolerance, unsigned  maxNormCycles, const std::vector< double > &  bandwidthValues, bool  useConvolve  )

protected

Constructor which explicitly specifies the complete set of bandwidth values to use in cross-validation

Member Function Documentation

bandwidthValues()

const std::vector<double>& npstat::AbsCVCopulaSmoother::bandwidthValues ( ) const

inline

Bandwidth values to cross-validate

getNFilters()

unsigned npstat::AbsCVCopulaSmoother::getNFilters ( ) const

inline

Number of bandwidth values to cross-validate

isConvolving()

bool npstat::AbsCVCopulaSmoother::isConvolving ( ) const

inline

Check how the kernel is used

lastCVValues()

const std::vector<double>& npstat::AbsCVCopulaSmoother::lastCVValues ( ) const

inline

Calculated values of the cross-validation criterion

lastFilterChosen()

unsigned npstat::AbsCVCopulaSmoother::lastFilterChosen ( ) const

inline

Index of the bandwidth best according to cross-validation

lastRegularizedFractions()

const std::vector<double>& npstat::AbsCVCopulaSmoother::lastRegularizedFractions ( ) const

inline

Fraction of bins that was affected by regularization

setConvolving()

void npstat::AbsCVCopulaSmoother::setConvolving ( const bool  b )

inline

Use either "filter" (kernel placement at the points in which the density is estimated) or "convolve" mode (kernel placement at the sample points)

---

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

• stat/AbsCVCopulaSmoother.hh