weka.experiment
Class PairedStatsCorrected

java.lang.Object
  weka.experiment.PairedStats
      weka.experiment.PairedStatsCorrected

public class PairedStatsCorrected

extends PairedStats

A class for storing stats on a paired comparison. This version is based on the corrected resampled t-test statistic, which uses the ratio of the number of test examples/the number of training examples.

For more information see:

Claude Nadeau and Yoshua Bengio, "Inference for the Generalization Error," Machine Learning, 2001.

Version:

$Revision: 1.2 $

Author:

Richard Kirkby (rkirkby@cs.waikato.ac.nz)

Field Summary

protected double

m_testTrainRatio The ratio used to correct the significane test

Fields inherited from class weka.experiment.PairedStats

correlation, count, differencesProbability, differencesSignificance, differencesStats, sigLevel, xStats, xySum, yStats

Constructor Summary

PairedStatsCorrected(double sig, double testTrainRatio)
Creates a new PairedStatsCorrected object with the supplied significance level and train/test ratio.

Method Summary

void	calculateDerived() Calculates the derived statistics (significance etc).

Methods inherited from class weka.experiment.PairedStats

add, main, subtract, toString

Methods inherited from class java.lang.Object

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

Field Detail

m_testTrainRatio

protected double m_testTrainRatio

The ratio used to correct the significane test

Constructor Detail

PairedStatsCorrected

public PairedStatsCorrected(double sig,
                            double testTrainRatio)

Creates a new PairedStatsCorrected object with the supplied significance level and train/test ratio.

Parameters:

sig - the significance level for comparisons

Method Detail

calculateDerived

public void calculateDerived()

Calculates the derived statistics (significance etc).

Overrides:

calculateDerived in class PairedStats