RWRemovePredictors< T, S > Class Template Reference

Holds information about the removal of predictor variables from a parameter calculation. More...

#include <rw/analytics/regcalc.h>

Inheritance diagram for RWRemovePredictors< T, S >:

Public Member Functions
	RWRemovePredictors ()
	RWRemovePredictors (const RWRemovePredictors< T, S > &r)
	RWRemovePredictors (const RWGenMat< T > &r, const RWMathVec< S > o, size_t startIndex, size_t numRemoved, bool setBaseToLast=false)
size_t	numRemoved () const
void	setNumRemoved (size_t nr)
void	setStartIndex (size_t i)
size_t	startIndex () const
Public Member Functions inherited from RWRegressionDataChange< T, S >
bool	baseToLast () const
const RWMathVec< S > &	newObservationVec () const
const RWGenMat< T > &	newRegressionMat () const
void	setBaseToLast (bool s)
void	setNewObservationVec (const RWMathVec< S > &o)
void	setNewRegressionMat (const RWGenMat< T > &r)

Detailed Description

template<class T, class S>
class RWRemovePredictors< T, S >

Class RWRemovePredictors contains information regarding the removal of predictor variables from a base parameter calculation object's base calculation. The removal information must consist of the following:

• The new regression matrix. This is the regression matrix used in the parameter calculation object's base calculation with the predictors removed.
• The new observation vector. This is assumed to be the same as the observation vector used in the parameter calculation object's base calculation.
• The predictors removed in the form of a starting index and number to remove.

Synopsis

#include <rw/analytics/regcalc.h>
RWRemovePredictors<double,double> a;

Example

#include <rw/analytics/lsqqr.h>
#include <rw/rstream.h>

int main()
{
  // Full model regression matrix.
  RWGenMat<double> regMat( "6x3 [1  1.3 .54 \
                                 1  3.5 .65 \
                                 1 -2.3 .88 \
                                 1  8.2 .76 \
                                 1 -4.2 .32 \
                                 1  2.4 .43]" );
  // Reduced model regression matrix. Contains the first two
  // columns of regMat.
  RWGenMat<double> reducedRegMat( regMat(RWAll,RWSlice(0,2)) );

  RWMathVec<double> observationVec( "[12.3 15.6 22.3
                                      44.1 32.5 65.2]" );

  RWLeastSqQRCalc calcObject;
  // Set the base calculation for the parameter calculation
  // object to the full regression matrix.
  calcObject.setBaseCalc( regMat, observationVec );
  if ( calcObject.fail() )
  {
    cout << "Parameter calculation for full model failed" <<
            endl;
    return 1;
  }
  cout << "Parameters for full model:" <<
           calcObject.parameters() << endl;

  // Remove the predictor data contained in the last column of
  // the full model regression matrix using the data change
  // class RWRemovePredictors. The column to be removed is
  // indicated as "beginning at index 2, remove 1 column."
   RWRemovePredictors<double,double> dataChange( reducedRegMat,
                                         observationVec, 2, 1 );
  calcObject.removePredFromBaseCalc( dataChange );

  if ( calcObject.fail() )
  {
    cout << "Parameter calculation for reduced model failed" <<
            endl;
    return 1;
  }
  cout << "Parameters for reduced model:" <<
          calcObject.parameters() << endl;
  return 0;
}

Constructor & Destructor Documentation

template<class T, class S>

RWRemovePredictors< T, S >::RWRemovePredictors ( )

inline

Default constructor. Constructs an empty RWRemovePredictors object.

template<class T, class S>

RWRemovePredictors< T, S >::RWRemovePredictors ( const RWRemovePredictors< T, S > &  r )

inline

Copy constructor. Constructs a copy of r.

template<class T, class S>

RWRemovePredictors< T, S >::RWRemovePredictors ( const RWGenMat< T > &  r, const RWMathVec< S >  o, size_t  startIndex, size_t  numRemoved, bool  setBaseToLast = false  )

inline

Constructs an RWRemovePredictors object with the given regression matrix and observation vector. It assumes that the matrix r is obtained from the base calculations regression matrix by deleting numRemoved columns beginning with index startIndex, and that the observation vector o is identical to the base calculations. If setBaseToLast is true, the base calculations regression matrix is set to r.

Member Function Documentation

template<class T, class S>

size_t RWRemovePredictors< T, S >::numRemoved ( ) const

inline

Returns the number of predictors removed.

template<class T, class S>

void RWRemovePredictors< T, S >::setNumRemoved ( size_t  nr )

inline

Sets the number of predictors to be removed.

template<class T, class S>

void RWRemovePredictors< T, S >::setStartIndex ( size_t  i )

inline

Sets the starting index.

template<class T, class S>

size_t RWRemovePredictors< T, S >::startIndex ( ) const

inline

Returns the starting index.