Container of multivariate statistics. More...

#include <MultivariateStatistics.h>

Collaboration diagram for Isis::MultivariateStatistics:

Public Member Functions
	MultivariateStatistics ()
	Constructs a Multivariate Statistics object with accumulators and counters set to zero.

	MultivariateStatistics (const PvlObject &inStats)
	Constructs a MulitvariateStatistics object from a PvlObject.

	~MultivariateStatistics ()
	Destructs a MultivariateStatistics object.

void	Reset ()
	Resets all accumulators to zero.

void	AddData (const double x, const double y, const unsigned int count)
	Add two arrays of doubles to the accumulators and counters.

void	AddData (double x, double y, unsigned int count=1)
	Add an x,y value to the accumulators and counters count times.

void	RemoveData (const double x, const double y, const unsigned int count)
	Remove an array of doubles from the accumulators and counters.

Isis::Statistics	X () const
	Returns a Stats object for all of the X data fed through the AddData method.

Isis::Statistics	Y () const
	Returns a Stats object for all of the Y data fed through the AddData method.

double	SumXY () const
	Returns the sum of x*y for all data given through the AddData method.

double	Covariance () const
	Computes and returns the covariance between the two data sets If there are no valid data (pixels) then NULL8 is returned.

double	Correlation () const
	Computes and returns the coefficient of correlation (between -1.0 and 1.0) of the two data sets.

void	LinearRegression (double &a, double &b) const
	Fits a line.

BigInt	ValidPixels () const
	Returns the number of valid pixels processed.

BigInt	InvalidPixels () const
	Returns the number of invalid pixels encountered.

BigInt	TotalPixels () const
	Returns the total number of pixels processed.

PvlObject	toPvl (QString name="MultivariateStatistics") const
	Serializes a multivariate statistics object as a PvlObject.

Private Member Functions
void	fromPvl (const PvlObject &inStats)
	Unserializes a multivariate statistics object from a PvlObject.

Private Attributes
Isis::Statistics	p_x
	A Statistics object holding x data.

Isis::Statistics	p_y
	A Statistics object holding y data.

double	p_sumxy
	The sum of x and y.

BigInt	p_validPixels
	The number of valid (computed) pixels.

BigInt	p_invalidPixels
	The number of invalid (ignored) pixels.

BigInt	p_totalPixels
	The total number of pixels (invalid and valid).

Detailed Description

Container of multivariate statistics.

This class is used to accumulate multivariate statisics on two double arrays. In particular, it is highly useful for obtaining the covariance, correlation, and linear regression analysis on the the data. It ignores input values which are Isis special pixel values. That is, if either co-aligned double value is a special pixel then both values are not used in any statistical computation.

See also: Histogram; Stats

Author: 2004-06-08 Jeff Anderson

History

2005-03-28 Leah Dahmer modified file to support Doxygen documentation.

2005-05-23 Jeff Anderson - Added 2GB+ file support

2012-01-03 Steven Lambright - Added AddData(double, double, unsigned int) for a significant performance improvement and to increase the consistency in the API relative to the Statistics class.

2016-07-15 Ian Humphrey - Added constructor to initialize a MultivariateStatistics object from a PvlObject. Added fromPvl() and toPvl() methods to allow for serialization/unserialization with PvlObjects. Updated unit test. References #2282.

Todo

This class needs an example.

For the below methods we will need to compute log x, loy y, sumx3, sumx4,sumx2y: void ExponentialRegression (double &a, double &b) const; void PowerRegression (double &a, double &b) const; void parabolicRegression (double &a, double &b, double &c);

Definition at line 54 of file MultivariateStatistics.h.

Constructor & Destructor Documentation

◆ MultivariateStatistics() [1/2]

Isis::MultivariateStatistics::MultivariateStatistics ( )

Constructs a Multivariate Statistics object with accumulators and counters set to zero.

Definition at line 25 of file MultivariateStatistics.cpp.

References Reset().

◆ MultivariateStatistics() [2/2]

Isis::MultivariateStatistics::MultivariateStatistics ( const PvlObject & inStats )

Constructs a MulitvariateStatistics object from a PvlObject.

Parameters

const PvlObject & - Input multivariate statistics.

Definition at line 35 of file MultivariateStatistics.cpp.

References fromPvl(), and Reset().

◆ ~MultivariateStatistics()

Isis::MultivariateStatistics::~MultivariateStatistics ( )

Destructs a MultivariateStatistics object.

Definition at line 54 of file MultivariateStatistics.cpp.

Member Function Documentation

◆ AddData() [1/2]

void Isis::MultivariateStatistics::AddData	(	const double *	x,
		const double *	y,
		const unsigned int	count )

Add two arrays of doubles to the accumulators and counters.

This method can be invoked multiple times, for example, once for each line in a cube, before obtaining statistics.

Parameters

x	Array of doubles to add.
y	Array of doubles to add.
count	Number of doubles to process.

Definition at line 66 of file MultivariateStatistics.cpp.

References Isis::Statistics::AddData(), Isis::IsValidPixel(), p_invalidPixels, p_sumxy, p_totalPixels, p_validPixels, p_x, and p_y.

Referenced by Isis::MaximumCorrelation::MatchAlgorithm().

◆ AddData() [2/2]

void Isis::MultivariateStatistics::AddData	(	double	x,
		double	y,
		unsigned int	count = 1 )

Add an x,y value to the accumulators and counters count times.

This method can be invoked multiple times before obtaining statistics.

Parameters

x	x value to add
y	y value to add
count	Number of times to add this x,y value

Definition at line 94 of file MultivariateStatistics.cpp.

References Isis::Statistics::AddData(), Isis::IsValidPixel(), p_invalidPixels, p_sumxy, p_totalPixels, p_validPixels, p_x, and p_y.

◆ Correlation()

double Isis::MultivariateStatistics::Correlation ( ) const

Computes and returns the coefficient of correlation (between -1.0 and 1.0) of the two data sets.

This can be used as a goodness-of-fit measurement. The close the correlation is two -1.0 or 1.0 the more likely the data sets are related (and therefore the regression equation is valid).

Returns: The coefficient of correlation. (between -1.0 and 1.0) The closer to 0.0 implies there is less correlation between the data sets. Returns NULL8 if correlation couldn't be computed.

Definition at line 171 of file MultivariateStatistics.cpp.

References Covariance(), p_validPixels, p_x, p_y, and Isis::Statistics::StandardDeviation().

Referenced by toPvl().