DataClipper Class Reference

A DataClipper gets a bunch of data and determines at what value to clip if a certain clip percentage is desired. More...

Public Member Functions
	DataClipper ()
bool	isEmpty () const
void	setApproxNrValues (od_int64 nrsamples, int statsize=2000)
void	putData (float)
void	putData (const float *, od_int64 sz)
void	putData (const ValueSeries< float > &, od_int64 sz)
void	putData (const ArrayND< float > &)
bool	calculateRange (float cliprate, Interval< float > &)
bool	calculateRange (float lowcliprate, float highcliprate, Interval< float > &)
bool	fullSort ()
bool	getRange (float cliprate, Interval< float > &) const
bool	getRange (float lowcliprate, float highcliprate, Interval< float > &) const
bool	getSymmetricRange (float cliprate, float midval, Interval< float > &) const
void	reset ()
const LargeValVec< float > &	statPts () const

Static Public Member Functions
static bool	calculateRange (float *vals, od_int64 nrvals, float lowcliprate, float highcliprate, Interval< float > &)

Protected Attributes
int	approxstatsize_
float	sampleprob_
bool	subselect_
LargeValVec< float >	samples_
Interval< float >	absoluterg_

Detailed Description

A DataClipper gets a bunch of data and determines at what value to clip if a certain clip percentage is desired.

For simple cases, where no subselection is needed (i.e. the stats will be performed on all values, and only one dataset is used) the static function calculateRange is good enough:

TypeSet<float> mydata;
Interval<float> range;
DataClipper::calculateRange( mydata.arr(), mydata.size(), 0.05, 0.05,
range );

If there are more than one dataset, or if a subselection is wanted, the class is used as follows:

1. Create object
2. If subselection is wanted, set total nr of samples and statsize with setApproxNrValues
3. Add all your sources putData
4. If you only want a fixed range, call calculateRange.
5. If you want to come back an get multiple ranges, call fullSort. After fullSort, the getRange functions can be called, any number of times.
6. To prepare the object for a new set of data, call reset.

Example

Array3D<float> somedata;
TypeSet<float> moredata;
float    otherdata;

DataClipper clipper;
setApproxNrValues( somedata.info().getTotalSz()+moredata.size()+1, 2000 );
clipper.putData( somedata );
clipper.putData( moredata.arr(), moredata.size() );
clipper.putData( otherdata );

clipper.fullSort();

Interval<float> clip99;
Interval<float> clip95;
Interval<float> clip90;
clipper.getRange( 0.01, clip99 );
clipper.getRange( 0.05, clip95 );
clipper.getRange( 0.10, clip90 );

Constructor & Destructor Documentation

DataClipper::DataClipper

(

)

cliprate is between 0 and 0.5, cliprate0 is the bottom cliprate, cliprate1 is the top cliprate, when cliprate1 is -1, it will get the value of cliprate0

Member Function Documentation

bool DataClipper::calculateRange	(	float	cliprate,
		Interval< float > &
	)

Does not do a full sort. Also performes reset

bool DataClipper::calculateRange	(	float	lowcliprate,
		float	highcliprate,
		Interval< float > &
	)

Does not do a full sort. Also performes reset

static bool DataClipper::calculateRange ( float *  vals, od_int64  nrvals, float  lowcliprate, float  highcliprate, Interval< float > &    )

static

Does not do a full sort.

Note

vals are modified.

bool DataClipper::fullSort

(

)

bool DataClipper::getRange	(	float	cliprate,
		Interval< float > &
	)		const

bool DataClipper::getRange	(	float	lowcliprate,
		float	highcliprate,
		Interval< float > &
	)		const

bool DataClipper::getSymmetricRange	(	float	cliprate,
		float	midval,
		Interval< float > &
	)		const

bool DataClipper::isEmpty ( ) const

inline

void DataClipper::putData

(

float

)

void DataClipper::putData	(	const float *	,
		od_int64	sz
	)

void DataClipper::putData	(	const ValueSeries< float > &	,
		od_int64	sz
	)

void DataClipper::putData

(

const ArrayND< float > &

)

void DataClipper::reset

(

)

void DataClipper::setApproxNrValues	(	od_int64	nrsamples,
		int	statsize = 2000
	)

Will make it faster if large amount of data is used. The Object will then randomly subselect on the input to get about statsize samples to do the stats on.

const LargeValVec<float>& DataClipper::statPts ( ) const

inline

Member Data Documentation

Interval<float> DataClipper::absoluterg_

protected

int DataClipper::approxstatsize_

protected

float DataClipper::sampleprob_

protected

LargeValVec<float> DataClipper::samples_

protected

bool DataClipper::subselect_

protected