org.broadinstitute.hellbender.utils

Class MannWhitneyU

java.lang.Object

org.broadinstitute.hellbender.utils.MannWhitneyU

public class MannWhitneyU
extends java.lang.Object

Imported with changes from Picard private.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	MannWhitneyU.RankedData The ranked data in one list and a list of the number of ties.
static class	MannWhitneyU.Result The results of performing a rank sum test.
static class	MannWhitneyU.TestStatistic The values of U1, U2 and the transformed number of ties needed for the calculation of sigma in the normal approximation.
static class	MannWhitneyU.TestType A variable that indicates if the test is one sided or two sided and if it's one sided which group is the dominator in the null hypothesis.

Field Summary

Fields

Modifier and Type	Field and Description
protected static org.apache.logging.log4j.Logger	logger

Constructor Summary

Constructors

Constructor and Description
MannWhitneyU()

Method Summary

Modifier and Type	Method and Description
MannWhitneyU.TestStatistic	calculateOneSidedU(double[] series1, double[] series2, MannWhitneyU.TestType whichSeriesDominates) Calculates the rank-sum test statisic U (sometimes W) from two sets of input data for a one-sided test with an int indicating which group is the dominator.
MannWhitneyU.RankedData	calculateRank(double[] series1, double[] series2)
MannWhitneyU.TestStatistic	calculateTwoSidedU(double[] series1, double[] series2) Calculates the two-sided rank-sum test statisic U (sometimes W) from two sets of input data.
MannWhitneyU.TestStatistic	calculateU1andU2(double[] series1, double[] series2) Rank both groups together and return a TestStatistic object that includes U1, U2 and number of ties for sigma
double	calculateZ(double u, int n1, int n2, double nties, MannWhitneyU.TestType whichSide) Calculates the Z score (i.e.
double	median(double[] data) Finds or calculates the median value of a sorted array of double.
double	permutationTest(double[] series1, double[] series2, double testStatU) Creates histogram of test statistics from a permutation test.
void	setMinimumSeriesLengthForNormalApproximation(int n) Sets the minimum number of values in each data series to use the normal distribution approximation.
MannWhitneyU.Result	test(double[] series1, double[] series2, MannWhitneyU.TestType whichSide) Constructs a new rank sum test with the given data.
double	transformTies(int numOfRanks, java.util.ArrayList<java.lang.Integer> numOfTies)

Methods inherited from class java.lang.Object

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

Field Detail

logger

protected static org.apache.logging.log4j.Logger logger

Constructor Detail

MannWhitneyU

public MannWhitneyU()

Method Detail

setMinimumSeriesLengthForNormalApproximation

public void setMinimumSeriesLengthForNormalApproximation(int n)

Sets the minimum number of values in each data series to use the normal distribution approximation.

calculateRank

public MannWhitneyU.RankedData calculateRank(double[] series1,
                                             double[] series2)

calculateU1andU2

public MannWhitneyU.TestStatistic calculateU1andU2(double[] series1,
                                                   double[] series2)

Rank both groups together and return a TestStatistic object that includes U1, U2 and number of ties for sigma

transformTies

public double transformTies(int numOfRanks,
                            java.util.ArrayList<java.lang.Integer> numOfTies)

calculateOneSidedU

public MannWhitneyU.TestStatistic calculateOneSidedU(double[] series1,
                                                     double[] series2,
                                                     MannWhitneyU.TestType whichSeriesDominates)

Calculates the rank-sum test statisic U (sometimes W) from two sets of input data for a one-sided test with an int indicating which group is the dominator. Returns a test statistic object with trueU and number of ties for sigma.

calculateTwoSidedU

public MannWhitneyU.TestStatistic calculateTwoSidedU(double[] series1,
                                                     double[] series2)

Calculates the two-sided rank-sum test statisic U (sometimes W) from two sets of input data. Returns a test statistic object with trueU and number of ties for sigma.

calculateZ

public double calculateZ(double u,
                         int n1,
                         int n2,
                         double nties,
                         MannWhitneyU.TestType whichSide)

Calculates the Z score (i.e. standard deviations from the mean) of the rank sum test statistics given input data of lengths n1 and n2 respectively, as well as the number of ties, for normal approximation only.

median

public double median(double[] data)

Finds or calculates the median value of a sorted array of double.

test

public MannWhitneyU.Result test(double[] series1,
                                double[] series2,
                                MannWhitneyU.TestType whichSide)

Constructs a new rank sum test with the given data.

Parameters:

series1 - group 1 data

series2 - group 2 data

whichSide - indicator of two sided test, 0 for two sided, 1 for series1 as dominator, 2 for series2 as dominator

Returns:

Result including U statistic, Z score, p-value, and difference in medians.

permutationTest

public double permutationTest(double[] series1,
                              double[] series2,
                              double testStatU)

Creates histogram of test statistics from a permutation test.

Parameters:

series1 - Data from group 1

series2 - Data from group 2

testStatU - Test statistic U from observed data

Returns:

P-value based on histogram with u calculated for every possible permutation of group tag.