RWLogisticFitAnalysis Class Reference

Calculates a number of goodness of fit quantities for a given RWLogisticRegression instance. More...

#include <rw/analytics/logfit.h>

Public Member Functions
	RWLogisticFitAnalysis ()
	RWLogisticFitAnalysis (const RWLogisticFitAnalysis &rhs)
	RWLogisticFitAnalysis (const RWLogisticRegression &reg)
	~RWLogisticFitAnalysis ()
double	deviance () const
double	GStatistic () const
double	GStatisticCriticalValue (double alpha=.05) const
int	GStatisticDegreesOfFreedom () const
double	GStatisticPValue () const
double	HLStatistic () const
double	HLStatisticCriticalValue (double alpha=.05) const
int	HLStatisticDegreesOfFreedom () const
const Histogram &	HLStatisticOutputHistogram () const
const Histogram &	HLStatisticPosObsHistogram () const
double	HLStatisticPValue () const
double	logLikelihood () const
size_t	modelOutputGroups () const
RWLogisticFitAnalysis &	operator= (const RWLogisticFitAnalysis &rhs)
double	PearsonStatistic () const
double	PearsonStatisticCriticalValue (double alpha=.05) const
int	PearsonStatisticDegreesOfFreedom () const
double	PearsonStatisticPValue () const
const RWMathVec< size_t >	predictorDataGroups () const
int	regressionDegreesOfFreedom () const
int	residualDegreesOfFreedom () const
void	setLogisticRegression (const RWLogisticRegression &lr)
void	setModelOutputGroups (size_t ngroups)
void	setPredictorDataGroups (const RWMathVec< size_t > &groups)

Detailed Description

Given RWLogisticRegression instance, class RWLogisticFitAnalysis calculates several goodness of fit quantities, including the G-statistic, Pearson statistic, and Hosmer and Lemeshow statistic, along with their P-values and critical values. These statistics are described in the Business Analysis Module User's Guide.

Synopsis

#include <rw/stats/logfit.h>
#include <rw/stats/logregress.h>
RWLogisticRegression lr;
RWLogisticFitAnalysis logfit(lr);

Example

#include <rw/analytics/logfit.h>
#include <rw/analytics/logregress.h>
#include <iostream>
 
int main() {
  RWGenMat<double> predData = "5x2 [1 234 2 431 3 333 4 654 5 788]";
  RWMathVec<bool> obsData(5, rwUninitialized);
  obsData[0] = obsData[3] = obsData[4] = true;
  obsData[1] = obsData[2] = false;
 
  // Create a logistic regression object.
  RWLogisticRegression model(predData, obsData);
 
  // Analyze the model; print out some summaries of the model's
  // goodness of fit.
  RWLogisticFitAnalysis logfit(model);
 
  std::cout << "Pearson statistic: " << logfit.PearsonStatistic()
            << std::endl;
  std::cout << "Pearson statistic P-value: "
            << logfit.PearsonStatisticPValue() << std::endl;
  std::cout << "Pearson statistic 10% critical value: "
            << logfit.PearsonStatisticCriticalValue(0.10) << std::endl;
  std::cout << "Pearson statistic degrees of freedom: "
            << logfit.PearsonStatisticDegreesOfFreedom() << std::endl;
  std::cout << "Predictor data groups for Pearson statistic: "
            << logfit.predictorDataGroups() << std::endl;
  std::cout << std::endl;
 
  std::cout << "HL statistic: " << logfit.HLStatistic() << std::endl;
  std::cout << "HL statistic P-value: " << logfit.HLStatisticPValue()
            << std::endl;
  std::cout << "HL statistic 10% critical value: "
            << logfit.HLStatisticCriticalValue(0.10) << std::endl;
  std::cout << "HL statistic degrees of freedom: "
            << logfit.HLStatisticDegreesOfFreedom() << std::endl;
  std::cout << "Bin counts for predictions: "
            << logfit.HLStatisticOutputHistogram() << std::endl;
  std::cout << "Bin counts for predictions associated\n"
            << "with positive observations: "
            << logfit.HLStatisticPosObsHistogram() << std::endl;
  std::cout << std::std::endl;
 
  return 0;
}

Constructor & Destructor Documentation

RWLogisticFitAnalysis() [1/3]

RWLogisticFitAnalysis::RWLogisticFitAnalysis

(

)

Constructs an empty logistic regression fit analysis object. Behavior is undefined.

RWLogisticFitAnalysis() [2/3]

RWLogisticFitAnalysis::RWLogisticFitAnalysis

(

const RWLogisticFitAnalysis &

rhs

)

Constructs a copy of rhs.

RWLogisticFitAnalysis() [3/3]

RWLogisticFitAnalysis::RWLogisticFitAnalysis

(

const RWLogisticRegression &

reg

)

Constructs an analysis object for the logistic regression object reg.

~RWLogisticFitAnalysis()

RWLogisticFitAnalysis::~RWLogisticFitAnalysis ( )

inline

Destructor.

Member Function Documentation

deviance()

double RWLogisticFitAnalysis::deviance ( ) const

inline

Returns the deviance from the saturated model.

GStatistic()

double RWLogisticFitAnalysis::GStatistic ( ) const

inline

Returns the G statistic value, described in the Business Analysis Module User's Guide, relative to the intercept only model.

GStatisticCriticalValue()

double RWLogisticFitAnalysis::GStatisticCriticalValue ( double alpha = .05 ) const

inline

Returns the critical value for the G statistic at the specified significance level.

GStatisticDegreesOfFreedom()

int RWLogisticFitAnalysis::GStatisticDegreesOfFreedom ( ) const

inline

Returns the degrees of freedom for the G statistic test value.

GStatisticPValue()

double RWLogisticFitAnalysis::GStatisticPValue ( ) const

inline

Returns the P-value for the G statistic according to a chi-square distribution.

HLStatistic()

double RWLogisticFitAnalysis::HLStatistic ( ) const

inline

Returns the Hosmer-Lemeshow test statistic, described in the Business Analysis Module User's Guide.

HLStatisticCriticalValue()

double RWLogisticFitAnalysis::HLStatisticCriticalValue ( double alpha = .05 ) const

inline

Returns the critical value for the Hosmer-Lemeshow statistic according to a chi-square distribution.

HLStatisticDegreesOfFreedom()

int RWLogisticFitAnalysis::HLStatisticDegreesOfFreedom ( ) const

inline

Returns degrees of freedom for the Hosmer-Lemeshow statistic test value.

HLStatisticOutputHistogram()

const Histogram & RWLogisticFitAnalysis::HLStatisticOutputHistogram ( ) const

inline

Returns the histogram of model predictions used in computing the Hosmer-Lemeshow statistic. This can be used to verify the exact bin boundaries, which can be critical to the statistic's value.

HLStatisticPosObsHistogram()

const Histogram & RWLogisticFitAnalysis::HLStatisticPosObsHistogram ( ) const

inline

Returns the histogram of model predictions whose corresponding observation value is positive. Seeing this histogram can be very useful for understanding the value for the statistic.

HLStatisticPValue()

double RWLogisticFitAnalysis::HLStatisticPValue ( ) const

inline

Returns the P-value for the Hosmer-Lemeshow statistic according to a chi-square distribution.

logLikelihood()

double RWLogisticFitAnalysis::logLikelihood ( ) const

inline

Returns log likelihood of the logistic regression model.

modelOutputGroups()

size_t RWLogisticFitAnalysis::modelOutputGroups ( ) const

inline

Returns the number of groups used for computing the Hosmer-Lemeshow statistic.

operator=()

RWLogisticFitAnalysis & RWLogisticFitAnalysis::operator=

(

const RWLogisticFitAnalysis &

rhs

)

Sets self equal to rhs.

PearsonStatistic()

double RWLogisticFitAnalysis::PearsonStatistic ( ) const

inline

Returns the Pearson test statistic. See the Business Analysis Module User's Guide for more information.

PearsonStatisticCriticalValue()

double RWLogisticFitAnalysis::PearsonStatisticCriticalValue ( double alpha = .05 ) const

inline

Returns the critical value for the Pearson statistic according to a chi-square distribution.

PearsonStatisticDegreesOfFreedom()

int RWLogisticFitAnalysis::PearsonStatisticDegreesOfFreedom ( ) const

inline

Returns degrees of freedom for the Pearson statistic test value.

PearsonStatisticPValue()

double RWLogisticFitAnalysis::PearsonStatisticPValue ( ) const

inline

Returns the P-value for the Pearson statistic according to a chi-square distribution.

predictorDataGroups()

const RWMathVec< size_t > RWLogisticFitAnalysis::predictorDataGroups ( ) const

inline

Returns the grouping used for the predictor matrix during computation of a Pearson statistic.

regressionDegreesOfFreedom()

int RWLogisticFitAnalysis::regressionDegreesOfFreedom ( ) const

inline

Returns the regression degrees of freedom for a regression model. This is defined as the number of predictor variables in the model.

residualDegreesOfFreedom()

int RWLogisticFitAnalysis::residualDegreesOfFreedom ( ) const

inline

Returns the residual degrees of freedom for a regression model. This is defined as the number of predictor patterns minus the number of parameters in the regression model.

setLogisticRegression()

void RWLogisticFitAnalysis::setLogisticRegression

(

const RWLogisticRegression &

lr

)

Specifies which regression model will be used for the fit analysis.

setModelOutputGroups()

void RWLogisticFitAnalysis::setModelOutputGroups

(

size_t

ngroups

)

Sets the grouping that should be used for the model's predicted values when computing a Hosmer-Lemeshow statistic. The model's predicted values are organized into a number of equal-mass bins; the number of bins is equal to ngroups.

setPredictorDataGroups()

void RWLogisticFitAnalysis::setPredictorDataGroups

(

const RWMathVec< size_t > &

groups

)

Sets the grouping that should be used for the predictor matrix when computing a Pearson statistic. The length of groups should equal the number of predictor variables in the model. The element groups[k] specifies the number of equally-spaced groups to be used for predictor variable k's data. Each element of groups must have a positive (nonzero) value.