weka.core
Class ContingencyTables

java.lang.Object
  weka.core.ContingencyTables

public class ContingencyTables

extends java.lang.Object

Class implementing some statistical routines for contingency tables.

Version:

$Revision: 1.5 $

Author:

Eibe Frank (eibe@cs.waikato.ac.nz)

Field Summary

private static double

log2 The natural logarithm of 2

Constructor Summary

ContingencyTables()

Method Summary

private static double	chiCell(double freq, double expected, boolean yates) Computes chi-value for one cell in a contingency table.
static double	chiSquared(double[][] matrix, boolean yates) Returns chi-squared probability for a given matrix.
static double	chiVal(double[][] matrix, boolean useYates) Computes chi-squared statistic for a contingency table.
static boolean	cochransCriterion(double[][] matrix) Tests if Cochran's criterion is fullfilled for the given contingency table.
static double	CramersV(double[][] matrix) Computes Cramer's V for a contingency table.
static double	entropy(double[] array) Computes the entropy of the given array.
static double	entropyConditionedOnColumns(double[][] matrix) Computes conditional entropy of the rows given the columns.
static double	entropyConditionedOnRows(double[][] matrix) Computes conditional entropy of the columns given the rows.
static double	entropyConditionedOnRows(double[][] train, double[][] test, double numClasses) Computes conditional entropy of the columns given the rows of the test matrix with respect to the train matrix.
static double	entropyOverColumns(double[][] matrix) Computes the columns' entropy for the given contingency table.
static double	entropyOverRows(double[][] matrix) Computes the rows' entropy for the given contingency table.
static double	gainRatio(double[][] matrix) Computes gain ratio for contingency table (split on rows).
private static double	lnFunc(double num) Help method for computing entropy.
static double	log2MultipleHypergeometric(double[][] matrix) Returns negative base 2 logarithm of multiple hypergeometric probability for a contingency table.
static void	main(java.lang.String[] ops) Main method for testing this class.
static double[][]	reduceMatrix(double[][] matrix) Reduces a matrix by deleting all zero rows and columns.
static double	symmetricalUncertainty(double[][] matrix) Calculates the symmetrical uncertainty for base 2.
static double	tauVal(double[][] matrix) Computes Goodman and Kruskal's tau-value for a contingency table.

Methods inherited from class java.lang.Object

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

Field Detail

log2

private static double log2

The natural logarithm of 2

Constructor Detail

ContingencyTables

public ContingencyTables()

Method Detail

chiSquared

public static double chiSquared(double[][] matrix,
                                boolean yates)

Returns chi-squared probability for a given matrix.

Parameters:

matrix - the contigency table

yates - is Yates' correction to be used?

Returns:

the chi-squared probability

chiVal

public static double chiVal(double[][] matrix,
                            boolean useYates)

Computes chi-squared statistic for a contingency table.

Parameters:

matrix - the contigency table

Returns:

the value of the chi-squared statistic

cochransCriterion

public static boolean cochransCriterion(double[][] matrix)

Tests if Cochran's criterion is fullfilled for the given contingency table. Rows and columns with all zeros are not considered relevant.

Parameters:

matrix - the contigency table to be tested

Returns:

true if contingency table is ok, false if not

CramersV

public static double CramersV(double[][] matrix)

Computes Cramer's V for a contingency table.

Parameters:

matrix - the contingency table

Returns:

Cramer's V

entropy

public static double entropy(double[] array)

Computes the entropy of the given array.

Parameters:

array - the array

Returns:

the entropy

entropyConditionedOnColumns

public static double entropyConditionedOnColumns(double[][] matrix)

Computes conditional entropy of the rows given the columns.

Parameters:

matrix - the contingency table

Returns:

the conditional entropy of the rows given the columns

entropyConditionedOnRows

public static double entropyConditionedOnRows(double[][] matrix)

Computes conditional entropy of the columns given the rows.

Parameters:

matrix - the contingency table

Returns:

the conditional entropy of the columns given the rows

entropyConditionedOnRows

public static double entropyConditionedOnRows(double[][] train,
                                              double[][] test,
                                              double numClasses)

Computes conditional entropy of the columns given the rows of the test matrix with respect to the train matrix. Uses a Laplace prior. Does NOT normalize the entropy.

Parameters:

train - the train matrix

test - the test matrix

Returns:

the entropy

entropyOverRows

public static double entropyOverRows(double[][] matrix)

Computes the rows' entropy for the given contingency table.

Parameters:

matrix - the contingency table

Returns:

the rows' entropy

entropyOverColumns

public static double entropyOverColumns(double[][] matrix)

Computes the columns' entropy for the given contingency table.

Parameters:

matrix - the contingency table

Returns:

the columns' entropy

gainRatio

public static double gainRatio(double[][] matrix)

Computes gain ratio for contingency table (split on rows). Returns Double.MAX_VALUE if the split entropy is 0.

Parameters:

matrix - the contingency table

Returns:

the gain ratio

log2MultipleHypergeometric

public static double log2MultipleHypergeometric(double[][] matrix)

Returns negative base 2 logarithm of multiple hypergeometric probability for a contingency table.

Parameters:

matrix - the contingency table

Returns:

the log of the hypergeometric probability of the contingency table

reduceMatrix

public static double[][] reduceMatrix(double[][] matrix)

Reduces a matrix by deleting all zero rows and columns.

Parameters:

matrix - the matrix to be reduced

symmetricalUncertainty

public static double symmetricalUncertainty(double[][] matrix)

Calculates the symmetrical uncertainty for base 2.

Parameters:

matrix - the contingency table

Returns:

the calculated symmetrical uncertainty

tauVal

public static double tauVal(double[][] matrix)

Computes Goodman and Kruskal's tau-value for a contingency table.

Parameters:

matrix - the contingency table

lnFunc

private static double lnFunc(double num)

Help method for computing entropy.

chiCell

private static double chiCell(double freq,
                              double expected,
                              boolean yates)

Computes chi-value for one cell in a contingency table.

Parameters:

freq - the observed frequency in the cell

expected - the expected frequency in the cell

Returns:

the chi-value for that cell; 0 if the expected value is too close to zero

main

public static void main(java.lang.String[] ops)

Main method for testing this class.