numRecipes

Class PoissonBinnedLikelihoodRatioFactory

java.lang.Object

numRecipes.PoissonBinnedLikelihoodRatioFactory

public class PoissonBinnedLikelihoodRatioFactory
extends java.lang.Object

Handle the ratio of the maximum likelihood ratio built from the binned Poisson. The null hypothesis (=background) and the alternative (=signal+background) hypothesis are describied, respectively, by the class PoissonBinnedLikelihoodCalculator.

    The null hypothesis        :  bg only
    The alternative hypothesis :  bg + lambda * signal
                                (lambda>=0)

    

Written originally by S. Yoshida for the IceCube EHE analysis. 2013/6/24

Field Summary

Fields

Modifier and Type	Field and Description
protected PoissonBinnedLikelihoodCalculator	calH0
protected PoissonBinnedLikelihoodCalculator	calH1
protected double	maxLogLikelihoodRatio maximized likelihood ratio log(L/L0)
protected double	poissonLogLikelihoodH0 log likelohood for the null hypothesis
protected double	poissonLogLikelihoodH1 log likelohood for the alternative hypothesis
protected double	signalFactorForMaximalLlhH1 the scale factor to maximize likelihood for the alternative hypothesis

Constructor Summary

Constructors

Constructor and Description
PoissonBinnedLikelihoodRatioFactory(PoissonBinnedLikelihoodCalculator calNull, PoissonBinnedLikelihoodCalculator calAlter) The default constructor.

Method Summary

Methods

Modifier and Type	Method and Description
void	calculateMaxLogLikelihoodRatio(double scaleFactor) calculate the maximized log likelihood ratio log(L/L0).
double	getMaxLogLikelihoodRatio(double scaleFactor, boolean runReplicaBGexperiment) return the maximized log likelihood ratio log(L/L0).
double	getPoissonLogLikelihoodH0(double scaleFactor) return the binned Poisson likelihood for the null (background-only) hypothesis
double	getPoissonLogLikelihoodH1(double scaleFactor) return the binned Poisson likelihood for the alternative (background + signal) hypothesis

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

calH0

protected PoissonBinnedLikelihoodCalculator calH0

calH1

protected PoissonBinnedLikelihoodCalculator calH1

poissonLogLikelihoodH0

protected double poissonLogLikelihoodH0

log likelohood for the null hypothesis

poissonLogLikelihoodH1

protected double poissonLogLikelihoodH1

log likelohood for the alternative hypothesis

maxLogLikelihoodRatio

protected double maxLogLikelihoodRatio

maximized likelihood ratio log(L/L0)

signalFactorForMaximalLlhH1

protected double signalFactorForMaximalLlhH1

the scale factor to maximize likelihood for the alternative hypothesis

Constructor Detail

PoissonBinnedLikelihoodRatioFactory

public PoissonBinnedLikelihoodRatioFactory(PoissonBinnedLikelihoodCalculator calNull,
                                   PoissonBinnedLikelihoodCalculator calAlter)

The default constructor. Sets the likelihhood calculator of the null and alternative hypotheses.

Method Detail

calculateMaxLogLikelihoodRatio

public void calculateMaxLogLikelihoodRatio(double scaleFactor)

calculate the maximized log likelihood ratio log(L/L0).

getPoissonLogLikelihoodH0

public double getPoissonLogLikelihoodH0(double scaleFactor)

return the binned Poisson likelihood for the null (background-only) hypothesis

       double scaleFactor:  scaleFactor whici is multiplied to mu_bg_i; if set to 1.0, no fudge factor is introduced.
       

getPoissonLogLikelihoodH1

public double getPoissonLogLikelihoodH1(double scaleFactor)

return the binned Poisson likelihood for the alternative (background + signal) hypothesis

       double scaleFactor:  scaleFactor whici is multiplied to mu_bg_i; if set to 1.0, no fudge factor is introduced.
       

getMaxLogLikelihoodRatio

public double getMaxLogLikelihoodRatio(double scaleFactor,
                              boolean runReplicaBGexperiment)

return the maximized log likelihood ratio log(L/L0). if runReplicaBGexperiment sets true, run the replica MC experiment with the H0 (=background only) hypothesis, and calculate the ratio aganst the results of the replica experiment. This method may be used for evaluating p-value for a given likelihood ratio (which can be provided by this method also setting runReplicaBGexperiment=false).

 
        boolean runReplicaBGexperiment:    true - run Replica experiment with null hypotheis
        double scaleFactor:  scaleFactor which is multiplied to mu_bg_i; if set to 1.0, no fudge factor is introduced.