Class Statistics

java.lang.Object
  |
  +--Statistics

class Statistics

extends java.lang.Object

The class contains various parametric and non-parametric probability statistics methods. These include: F-test, t-test, p-values for t-test, histogram, mean, stdDev, median, mode, extrema, etc.

NOTE: Statistics package (derived from WebGel and GELLAB-II which were derived from Numerical Recipes, etc.).

This work was produced by Peter Lemkin of the National Cancer Institute, an agency of the United States Government. As a work of the United States Government there is no associated copyright. It is offered as open source software under the Mozilla Public License (version 1.1) subject to the limitations noted in the accompanying LEGAL file. This notice must be included with the code. The MAExplorer Mozilla and Legal files are available on http://maexplorer.sourceforge.net/.

Version:

$Date: 2003/02/20 21:35:04 $ $Revision: $

Author:

P. Lemkin (NCI), G. Thornwall (SAIC), NCI-Frederick, Frederick, MD

See Also:

MAExplorer Home

Field Summary

(package private) float[][]	condData data[0:nConditions-1][sampleNbrInClass] for computations, set by calcMeanAndVariance()
(package private) float	deltaBinH width of the histogram bins computed as: nBinsH/(maxDataH-minDataH)
(package private) double	dF CALC: degrees of freedom of t-test or F-test (2 conditions) set by calcTandPvalues()
(package private) double	dfBetween CALC: degrees of freedom dfBetween, f-statistic nConditions test set by calcNCondFtestStat()
(package private) double	dFks CALC: degrees of freedom for KS-test set by calcKStestStat()
(package private) double	dfWithin CALC: degrees of freedom dfWithin, f-statistic nConditions test set by calcNCondFtestStat()
(package private) double	f CALC: calculated f statistic set by calcTandPvalues()
(package private) double	fStat CALC: f-statistic 2 conditions
(package private) double	fStatNconds CALC: f-statistic nConditions test set by calcNCondFtestStat()
(package private) int[]	hist histogram of size [0:nBinsH-1] set by calcHistStats()
(package private) double	ksD CALC: KS-test Kolmogorov-Smirnov D statistic set by calcKStestStat()
private MAExplorer	mae link to global instance of MAExplorer
(package private) float	maxDataH histogram data max value set by calcHistStats()
(package private) double[]	mean means[0:nConditions-1] of data for computations, set by calcMeanAndVariance()
(package private) float	meanAbsDevH histogram data mean absolute deviation set by calcHistStats()
(package private) float	meanH histogram data median set by calcHistStats()
(package private) int	meanIdx index of mean in hist[] set by calcHistStats()
(package private) float	medianH histogram data median set by calcHistStats()
(package private) int	medianIdx index of median in hist[] set by calcHistStats()
(package private) float	minDataH histogram data min value set by calcHistStats()
(package private) double	mnSqBetween CALC: mean square between variance, f-statistic nConditions test set by calcNCondFtestStat()
(package private) double	mnSqWithin CALC: mean square within variance, f-statistic nConditions test set by calcNCondFtestStat()
(package private) float	modeH histogram data mode set by calcHistStats()
(package private) int	modeIdx index of mode in hist[] set by calcHistStats()
(package private) int	nBinsH 0 if none.
private int	nCondAlloc size of in mean,variance,stdDev,condData,nCondData stats arrays set by calcNCondFtestStat()
(package private) int[]	nCondData data[0:nConditions-1][sampleNbrInClass] for computations, set by calcMeanAndVariance()
(package private) int	nConditions # of conditions in mean,var stats set by calcNCondFtestStat()
(package private) double	pF CALC: f-test p-value w/NULL hypoth set by calcTandPvalues()
(package private) double	pFnConds CALC: f-test p-value w/NULL hypoth for nConditions set by calcNCondFtestStat()
(package private) double	pKS CALC: KS-test p-value w/NULL hypoth set by calcKStestStat()
(package private) double	pT CALC: t-test p-value w/NULL hypoth set by calcTandPvalues()
(package private) double[]	stdDev stdDev[0:nConditions-1] of data for computations, set by calcMeanAndVariance() if calcStdDevFlag set on call
(package private) float	stdDevH histogram data stdDev set by calcHistStats()
(package private) double	t CALC: t or t' statistic set by calcTandPvalues()
(package private) java.lang.String	title for data used in histogram set by calcHistStats()
(package private) char	useTest CALC: 'B' or 'T' - t-test to use set by calcTandPvalues()
(package private) double[]	variance variance[0:nConditions-1] of data for computations, set by calcMeanAndVariance()

Constructor Summary

(package private)	Statistics() Statistics() - constructor
(package private)	Statistics(MAExplorer mae) Statistics() - constructor

Method Summary

(package private) boolean	calcFprobFromVariances(int n1, int n2, double var1, double var2) calcFprobFromVariances() - calculate 2-tailed f probility that variables are same.
(package private) int	calcHistStats(java.lang.String title, int nBins, float[] data, int nData) calcHistStats() - compute and analyze histogram generating statistics.
(package private) int	calcHistStats(java.lang.String title, int nBins, float[] data, int nData, int[] hist) calcHistStats() - compute and analyze a histogram for whatever range of data is given.
(package private) boolean	calcKStestStat(double[] data1, int n1, double[] data2, int n2) calcKStestStat() - calculate Kolmogorov-Smirnov ksD, pKS, dFks from (n1,data1) and (n2,data2).
(package private) boolean	calcMeanAndVariance(float[] dataS, int nSamples, int conditionK, boolean calcStdDevFlag) calcMeanAndVariance() - compute the mean and variance of dataS[] for 1 condition and save the results inmean[conditionK] and variance[conditionK] arrays.
(package private) boolean	calcNCondFtestStat(float[][] data, int[] nData, int nConditions) calcNCondFtestStat() - calc.
(package private) boolean	calcTandPvalues(int n1, int n2, double m1, double m2, double s1, double s2) calcTandPvalues() - calculate f, t, p, dF from (n1,m1,s1) and (n2,m2,s2).
(package private) boolean	calcWMWtestStat(int n1, int n2, double m1, double m2, double s1, double s2) calcWMWtestStat() - calculate WMW statistics from (n1,m1,s1) and (n2,m2,s2).

Methods inherited from class java.lang.Object

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

Field Detail

mae

private MAExplorer mae

link to global instance of MAExplorer

pT

double pT

CALC: t-test p-value w/NULL hypoth set by calcTandPvalues()

pF

double pF

CALC: f-test p-value w/NULL hypoth set by calcTandPvalues()

pKS

double pKS

CALC: KS-test p-value w/NULL hypoth set by calcKStestStat()

pFnConds

double pFnConds

CALC: f-test p-value w/NULL hypoth for nConditions set by calcNCondFtestStat()

f

double f

CALC: calculated f statistic set by calcTandPvalues()

t

double t

CALC: t or t' statistic set by calcTandPvalues()

fStat

double fStat

CALC: f-statistic 2 conditions

ksD

double ksD

CALC: KS-test Kolmogorov-Smirnov D statistic set by calcKStestStat()

dF

double dF

CALC: degrees of freedom of t-test or F-test (2 conditions) set by calcTandPvalues()

dFks

double dFks

CALC: degrees of freedom for KS-test set by calcKStestStat()

useTest

char useTest

CALC: 'B' or 'T' - t-test to use set by calcTandPvalues()

fStatNconds

double fStatNconds

CALC: f-statistic nConditions test set by calcNCondFtestStat()

mnSqWithin

double mnSqWithin

CALC: mean square within variance, f-statistic nConditions test set by calcNCondFtestStat()

mnSqBetween

double mnSqBetween

CALC: mean square between variance, f-statistic nConditions test set by calcNCondFtestStat()

dfWithin

double dfWithin

CALC: degrees of freedom dfWithin, f-statistic nConditions test set by calcNCondFtestStat()

dfBetween

double dfBetween

CALC: degrees of freedom dfBetween, f-statistic nConditions test set by calcNCondFtestStat()

title

java.lang.String title

for data used in histogram set by calcHistStats()

meanIdx

int meanIdx

index of mean in hist[] set by calcHistStats()

medianIdx

int medianIdx

index of median in hist[] set by calcHistStats()

modeIdx

int modeIdx

index of mode in hist[] set by calcHistStats()

nBinsH

int nBinsH

0 if none. size of histogram set by calcHistStats()

hist

int[] hist

histogram of size [0:nBinsH-1] set by calcHistStats()

medianH

float medianH

histogram data median set by calcHistStats()

modeH

float modeH

histogram data mode set by calcHistStats()

meanH

float meanH

histogram data median set by calcHistStats()

stdDevH

float stdDevH

histogram data stdDev set by calcHistStats()

meanAbsDevH

float meanAbsDevH

histogram data mean absolute deviation set by calcHistStats()

minDataH

float minDataH

histogram data min value set by calcHistStats()

maxDataH

float maxDataH

histogram data max value set by calcHistStats()

deltaBinH

float deltaBinH

width of the histogram bins computed as: nBinsH/(maxDataH-minDataH)

nCondAlloc

private int nCondAlloc

size of in mean,variance,stdDev,condData,nCondData stats arrays set by calcNCondFtestStat()

nConditions

int nConditions

# of conditions in mean,var stats set by calcNCondFtestStat()

condData

float[][] condData

data[0:nConditions-1][sampleNbrInClass] for computations, set by calcMeanAndVariance()

nCondData

int[] nCondData

data[0:nConditions-1][sampleNbrInClass] for computations, set by calcMeanAndVariance()

mean

double[] mean

means[0:nConditions-1] of data for computations, set by calcMeanAndVariance()

variance

double[] variance

variance[0:nConditions-1] of data for computations, set by calcMeanAndVariance()

stdDev

double[] stdDev

stdDev[0:nConditions-1] of data for computations, set by calcMeanAndVariance() if calcStdDevFlag set on call

Constructor Detail

Statistics

Statistics()

Statistics() - constructor

Statistics

Statistics(MAExplorer mae)

Statistics() - constructor

Parameters:

mae - is MAExplorer instance

Method Detail

calcMeanAndVariance

boolean calcMeanAndVariance(float[] dataS,
                            int nSamples,
                            int conditionK,
                            boolean calcStdDevFlag)

calcMeanAndVariance() - compute the mean and variance of dataS[] for 1 condition and save the results inmean[conditionK] and variance[conditionK] arrays. Also save the dataS in data[conditionK] and nSamples in nData[conditionK]. This method was derived Snedecore and Chochran Statistical Methods.

Parameters:

data - is array of size [0:nSamples-1] of data

nSamples - is size of data

conditionK - is the class # associated with this data (start at 0)

calcStdDevFlag - also compute stdDev[conditionK] as well

Returns:

true if succeed

calcFprobFromVariances

boolean calcFprobFromVariances(int n1,
                               int n2,
                               double var1,
                               double var2)

calcFprobFromVariances() - calculate 2-tailed f probility that variables are same. It computes:

    fStat - the f-statistic
    pF    - CALC: probab. vars. same

This method was derived from GELLAB-II which was derived from Numerical Recipes in C and Snedecore and Chochran Statistical Methods.

Parameters:

n1 - # samples class 1

n2 - # samples class 2

var1 - variance of class 1

var2 - variance of class 2

Returns:

true and set variables if succeed, else false if any problems.

See Also:

MathMAE.nr_betai(double, double, double)

calcTandPvalues

boolean calcTandPvalues(int n1,
                        int n2,
                        double m1,
                        double m2,
                        double s1,
                        double s2)

calcTandPvalues() - calculate f, t, p, dF from (n1,m1,s1) and (n2,m2,s2). Use Behrens-Fisher/Satterthwaite estimate for t and dF if f-stat is < 0.05 p-value that variances are different. Otherwise use the standard student t-statistic with DF= (n1+n2-2). If you want to force the test the set useTest to 'B' or 'T', else it will pick the test to use (ie. TB or TP). It uses the algorithm described Numerical Recipes in C (1st Ed) for estimating p-value given the t-statistic using the incomplete beta function betai(). It computes:

    f - calculated f statistic
    t - t or t' statistic
    pT - t-test p-value w/NULL hypoth
    pF - f-test p-value w/NULL hypoth
    dF - degrees of freedom
 

This method was derived from GELLAB-II which was derived from Numerical Recipes in C and Snedecore and Chochran Statistical Methods.

Parameters:

n1 - # samples in class 1

n2 - # samples in class 2

m1 - sample mean class 1

m2 - sample mean class 2

s1 - sample std dev class 1

s2 - sample std dev class 2

Returns:

false if any of the data is invalid (need >= 2 samples/class) or the beta fct fails.

See Also:

MathMAE.nr_betai(double, double, double), calcFprobFromVariances(int, int, double, double)

calcWMWtestStat

boolean calcWMWtestStat(int n1,
                        int n2,
                        double m1,
                        double m2,
                        double s1,
                        double s2)

calcWMWtestStat() - calculate WMW statistics from (n1,m1,s1) and (n2,m2,s2). If enabled, do WMW-test for 2 classes. [TODO] - not available yet

Parameters:

n1 - # samples in class 1

n2 - # samples in class 2

m1 - sample mean class 1

m2 - sample mean class 2

s1 - sample std dev class 1

s2 - sample std dev class 2

Returns:

false if any of the data is invalid (need >= 2 samples/class)

calcKStestStat

boolean calcKStestStat(double[] data1,
                       int n1,
                       double[] data2,
                       int n2)

calcKStestStat() - calculate Kolmogorov-Smirnov ksD, pKS, dFks from (n1,data1) and (n2,data2). DF= (n1+n2-2). It computes:

    ksD - D statistic
    pKS - KS test p-value w/NULL hypoth
    dFks - degrees of freedom
 

This method was derived from GELLAB-II which was derived from Numerical Recipes in C and Snedecore and Chochran Statistical Methods.

Parameters:

data1 - sample data class 1

n1 - # samples in class 1

data2 - sample data class 2

n2 - # samples in class 2

Returns:

false if any of the data is invalid (need >= 2 samples/class)

calcHistStats

int calcHistStats(java.lang.String title,
                  int nBins,
                  float[] data,
                  int nData)

calcHistStats() - compute and analyze histogram generating statistics. for whatever range of data is given. This computes and returns the results in variables of this instance: hist[0:nBins-1], medianH, modeH, meanH, stdDevH, meanAbsDevH, minDataH, maxDataH, deltaBinH

Parameters:

title - for data

nBins - size of hist[]

data - of size [0:nData-1]

nData - size of data array

Returns:

nBins if successful.

See Also:

calcHistStats(java.lang.String, int, float[], int)

calcHistStats

int calcHistStats(java.lang.String title,
                  int nBins,
                  float[] data,
                  int nData,
                  int[] hist)

calcHistStats() - compute and analyze a histogram for whatever range of data is given. This computes and returns the results in variables of this instance:

   hist[0:nBins-1], medianH, modeH, meanH,
   stdDevH, meanAbsDevH, minDataH, maxDataH, deltaBinH.

Note: if hist is int[nBins+1], then it will not be allocated.

Parameters:

title - for data

nBins - size of hist[]

data - of size [0:nData-1]

nData - size of data array

hist - opt. [nBins+1] else null in which case it will allocate it locally

Returns:

nBins if successful else 0.

calcNCondFtestStat

boolean calcNCondFtestStat(float[][] data,
                           int[] nData,
                           int nConditions)

calcNCondFtestStat() - calc. F-test statistics of data[0:nConditions-1][samples] It computes:

    pFnConds - p value
    fStatNconds - f statistic
    mnSqWithin - mean within class variance
    mnSqBetween - mean between class variance
    dF1 - degrees of freedom df1
    dF2 - degrees of freedom df2
 

This method was derived from GELLAB-II which was derived from Numerical Recipes in C, 2nd Edition, pg 619, Sec. 14.2, and Snedecore and Chochran Statistical Methods.

Parameters:

data - sample data[nConditions][sampleNbrInCondition]

nData - # samples in each [nConditions]

nConditions - # of conditions

Returns:

false if any of the data is invalid (need >1 sample/Condition)