#include <vcl_iterator.h>

Functions
template<class O , class RanIter >
O	rrel_util_median_abs_dev_scale (const RanIter &begin, const RanIter &end, int dof, O *)
	Compute the median absolute deviation (MAD) scale estimate of the residuals.
template<class T >
double	rrel_util_median_abs_dev_scale (const T &begin, const T &end, int dof=1)
	Convenience function.
template<class O , class InpIter >
O	rrel_util_weighted_scale (const InpIter &residuals_first, const InpIter &residuals_end, const InpIter &weights_first, int dof, O *)
	Compute the scale using robustly weighted residuals.
template<class InpIter >
double	rrel_util_weighted_scale (const InpIter &residuals_first, const InpIter &residuals_end, const InpIter &weights_first, int dof=1)
	Convenience function.
template<class T , class RanIter >
void	rrel_util_median_and_scale (RanIter first, RanIter last, T &median, T &scale, int dof=1)
	Compute the median and the scale (relative to the median).
template<class T , class InpIter >
void	rrel_util_median_and_scale_copy (InpIter first, InpIter last, T &median, T &scale, int dof=1)
	Compute the median and the scale (relative to the median).
template<class T , class RanIter >
void	rrel_util_intercept_adjustment (RanIter first, RanIter last, T &center, T &half_width, int dof=1)
	Compute the center and half width of the narrowest interval containing half the points in the residuals.
template<class T , class InpIter >
void	rrel_util_intercept_adjustment_copy (InpIter first, InpIter last, T &center, T &half_width, int dof=1)
	Compute the center and half width of the narrowest interval containing half the points in the residuals.
template<class T , class RanIter >
void	rrel_util_intercept_adjust_stats (RanIter first, RanIter last, T &robust_mean, T &robust_std, T &inlier_frac, int dof=1)
	Use the intercept adjustment technique to estimate the robust mean, standard deviation, and inlier fraction.
template<class T , class InpIter >
void	rrel_util_intercept_adjust_stats_copy (InpIter first, InpIter last, T &robust_mean, T &robust_std, T &inlier_frac, int dof=1)
	Use the intercept adjustment technique to estimate the robust mean, standard deviation, and inlier fraction.

Detailed Description

Author:: Chuck Stewart

Date:: Summer 2001

Definition in file rrel_util.h.

Function Documentation

template<class T , class RanIter >

void rrel_util_intercept_adjust_stats	(	RanIter	first,
		RanIter	last,
		T &	robust_mean,
		T &	robust_std,
		T &	inlier_frac,
		int	dof = `1`
	)

Use the intercept adjustment technique to estimate the robust mean, standard deviation, and inlier fraction.

template<class T , class InpIter >

void rrel_util_intercept_adjust_stats_copy	(	InpIter	first,
		InpIter	last,
		T &	robust_mean,
		T &	robust_std,
		T &	inlier_frac,
		int	dof = `1`
	)

Use the intercept adjustment technique to estimate the robust mean, standard deviation, and inlier fraction.

Same as rrel_util_intercept_adjustment_stats(), except that the input is copied first, and so is unchanged.

Definition at line 174 of file rrel_util.txx.

template<class T , class RanIter >

void rrel_util_intercept_adjustment	(	RanIter	first,
		RanIter	last,
		T &	center,
		T &	half_width,
		int	dof = `1`
	)

Compute the center and half width of the narrowest interval containing half the points in the residuals.

Find the center and half width of the narrowest interval containing half the points in the input set of residuals, accessed through the iterators first and last. This is the "intercept adjustment" technique of Rousseeuw (J. Amer. Stat. Assoc. 1984). The order of the residuals is destroyed and in fact, as a side effect, the residuals are sorted. The residuals are assumed to be signed values.

template<class T , class InpIter >

void rrel_util_intercept_adjustment_copy	(	InpIter	first,
		InpIter	last,
		T &	center,
		T &	half_width,
		int	dof = `1`
	)

Compute the center and half width of the narrowest interval containing half the points in the residuals.

Same as rrel_util_intercept_adjustment(), except that the input is copied first, and so is unchanged.

Definition at line 123 of file rrel_util.txx.

template<class O , class RanIter >

O rrel_util_median_abs_dev_scale	(	const RanIter &	begin,
		const RanIter &	end,
		int	dof,
		O *
	)

Compute the median absolute deviation (MAD) scale estimate of the residuals.

Find the median absolve deviation (MAD) scale estimate of a set of residuals. Assume a Gaussian target distribution. The input residuals may be signed or unsigned values, but they are converted to unsigned. The order of the residuals is destroyed.

The iterators should be random access iterators.

The dummy parameter is used to determine the return type:

    float s = rrel_util_median_abs_dev_scale( v.begin(), v.end(), 1, (float*)0 );

template<class T >

double rrel_util_median_abs_dev_scale	(	const T &	begin,
		const T &	end,
		int	dof = `1`
	)

Convenience function.

Calls the other rrel_util_median_abs_dev_scale() with a return type of double.

Definition at line 35 of file rrel_util.h.

template<class T , class RanIter >

void rrel_util_median_and_scale	(	RanIter	first,
		RanIter	last,
		T &	median,
		T &	scale,
		int	dof = `1`
	)

Compute the median and the scale (relative to the median).

Find the median and then, using the median of absolute deviations from the median, the scale. Assume a Gaussian target distribution. Both the order and the values themselves are changed. Random access iterators are required.

template<class T , class InpIter >

void rrel_util_median_and_scale_copy	(	InpIter	first,
		InpIter	last,
		T &	median,
		T &	scale,
		int	dof = `1`
	)

Compute the median and the scale (relative to the median).

Same as rrel_util_median_and_scale(), except that the input is copied first, and so is unchanged.

Definition at line 80 of file rrel_util.txx.

template<class O , class InpIter >

O rrel_util_weighted_scale	(	const InpIter &	residuals_first,
		const InpIter &	residuals_end,
		const InpIter &	weights_first,
		int	dof,
		O *
	)

Compute the scale using robustly weighted residuals.

Compute the scale using robustly weighted residuals. The dummy parameter is used to determine the return type. Internal calculations are also done with this type.

The iterators should be input iterators.

    float s = rrel_util_weighted_scale( r.begin(), r.end(),
                                        w.begin(), w.end(), 1, (float*)0 );

template<class InpIter >

double rrel_util_weighted_scale	(	const InpIter &	residuals_first,
		const InpIter &	residuals_end,
		const InpIter &	weights_first,
		int	dof = `1`
	)		`[inline]`

Convenience function.

Calls the other rrel_util_weighted_scale() with a return type of double.

Definition at line 66 of file rrel_util.h.

Functions

Detailed Description

Function Documentation