org.dkpro.statistics.significance

Class Significance

java.lang.Object

org.dkpro.statistics.significance.Significance

public class Significance
extends Object

Computes statistical significance.

Author:

zesch

Constructor Summary

Constructors

Constructor and Description
Significance()

Method Summary

Modifier and Type	Method and Description
static double	getSignificance(double[] sample1, double[] sample2) Uses a paired t-test to test whether the correlation value computed from these datasets is significant.
static double	getSignificance(double correlation1, double correlation2, int n1, int n2) Computes the significance of the difference between two correlations.
static double	getSignificance(List<Double> list1, List<Double> list2)
static boolean	testCorrelations(double correlation1, double correlation2, int n1, int n2, double alpha) Tests two correlation values for equality.

Methods inherited from class java.lang.Object

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

Constructor Detail

Significance

public Significance()

Method Detail

testCorrelations

public static boolean testCorrelations(double correlation1,
                                       double correlation2,
                                       int n1,
                                       int n2,
                                       double alpha)
                                throws org.apache.commons.math.MathException

Tests two correlation values for equality. See page 745 Press, W. H. (2007). Numerical Recipes 3rd Edition: The Art of Scientific Computing. Cambridge University Press. Null-Hypothesis: Both samples of pairs show the same correlation strength, i.e., r1 = r2. Assumptions: The values of both members of both samples of pairs are Normal (bivariate) distributed. Should only be used when both samples (n1 and n2) are larger than 10. Scale: Interval (for the raw data). Procedure: The two correlation coefficients are transformed with the Fisher Z-transformation: Zf = 1/2 * ln( (1+r) / (1-r) ) The difference is z = (Zf1 - Zf2) / ( SQRT( 2 ) * SQRT( 1/(N1-3) + 1/(N2-3) ) ) p-value is computed using the complementary error function p = 2 * (1 - pnorm(abs(z))) Uses: - for testing whether to correlations differ significantly

Parameters:

correlation1 - The first correlation value.

correlation2 - The second correlation value.

n1 - The size of the first dataset.

n2 - The size of the second dataset.

alpha - The significance level.

Returns:

True, if the null hypothesis is rejected, i.e. the difference between the correlation values is significant. False otherwise.

Throws:

org.apache.commons.math.MathException

getSignificance

public static double getSignificance(double[] sample1,
                                     double[] sample2)
                              throws IllegalArgumentException,
                                     org.apache.commons.math.MathException

Uses a paired t-test to test whether the correlation value computed from these datasets is significant.

Parameters:

sample1 - The first dataset vector.

sample2 - The second dataset vector.

Returns:

The significance value p.

Throws:

org.apache.commons.math.MathException

IllegalArgumentException

getSignificance

public static double getSignificance(List<Double> list1,
                                     List<Double> list2)
                              throws IllegalArgumentException,
                                     org.apache.commons.math.MathException

Throws:

org.apache.commons.math.MathException

IllegalArgumentException

See Also:

org.tud.sir.util.statistics.Significance#getSignificance(double[], double[])

getSignificance

public static double getSignificance(double correlation1,
                                     double correlation2,
                                     int n1,
                                     int n2)
                              throws org.apache.commons.math.MathException

Computes the significance of the difference between two correlations.

Throws:

org.apache.commons.math.MathException

See Also:

org.tud.sir.util.statistics.Significance.testCorrelations