Example usage for org.apache.commons.math3.distribution FDistribution FDistribution

List of usage examples for org.apache.commons.math3.distribution FDistribution FDistribution

  1. HOME
  2. Java
  3. org.apache.commons
  4. org.apache.commons.math3.distribution.*
  5. FDistribution
  6. FDistribution

Introduction

In this page you can find the example usage for org.apache.commons.math3.distribution FDistribution FDistribution.

Prototype

public FDistribution(double numeratorDegreesOfFreedom, double denominatorDegreesOfFreedom)
        throws NotStrictlyPositiveException

Source Link

Document

Creates an F distribution using the given degrees of freedom.

Usage

From source file:dase.timeseries.analysis.GrangerTest.java

/**
 * Returns p-value for Granger causality test.
 *
 * @param y/*  w  ww  . j  a  v  a 2 s  .  c o m*/
 *            - predictable variable
 * @param x
 *            - predictor
 * @param L
 *            - lag, should be 1 or greater.
 * @return p-value of Granger causality
 */
public static double granger(double[] y, double[] x, int L) {
    OLSMultipleLinearRegression h0 = new OLSMultipleLinearRegression();
    OLSMultipleLinearRegression h1 = new OLSMultipleLinearRegression();

    double[][] laggedY = createLaggedSide(L, y);

    double[][] laggedXY = createLaggedSide(L, x, y);

    int n = laggedY.length;

    h0.newSampleData(strip(L, y), laggedY);
    h1.newSampleData(strip(L, y), laggedXY);

    double rs0[] = h0.estimateResiduals();
    double rs1[] = h1.estimateResiduals();

    double RSS0 = sqrSum(rs0);
    double RSS1 = sqrSum(rs1);

    double ftest = ((RSS0 - RSS1) / L) / (RSS1 / (n - 2 * L - 1));

    System.out.println(RSS0 + " " + RSS1);
    System.out.println("F-test " + ftest);

    FDistribution fDist = new FDistribution(L, n - 2 * L - 1);

    double pValue = 1.0 - fDist.cumulativeProbability(ftest);
    System.out.println("P-value " + pValue);
    return pValue;
}

From source file:com.itemanalysis.psychometrics.reliability.AbstractScoreReliability.java

public double[] confidenceInterval() {
    double numberOfExaminees = matrix.getMaxSampleSize();
    double[] confidenceInterval = new double[2];
    double numberOfItems = (double) nItems;
    double df1 = numberOfExaminees - 1.0;
    double df2 = (numberOfExaminees - 1.0) * (numberOfItems - 1.0);
    FDistribution fDist = new FDistribution(df1, df2);
    try {//  w  w  w . j  av a  2 s. c  o m
        confidenceInterval[0] = 1.0 - ((1.0 - this.value()) * fDist.inverseCumulativeProbability(0.975));
        confidenceInterval[1] = 1.0 - ((1.0 - this.value()) * fDist.inverseCumulativeProbability(0.025));
    } catch (Exception ex) {
        confidenceInterval[0] = Double.NaN;
        confidenceInterval[1] = Double.NaN;
    }
    return confidenceInterval;
}

From source file:com.itemanalysis.psychometrics.reliability.ReliabilityInterval.java

public double[] confidenceInterval() {
    double N = sampleSize;
    double nI = numberOfVariables;
    double df1 = N - 1.0;
    double df2 = (N - 1.0) * (nI - 1.0);
    double[] ci = new double[2];
    FDistribution fDist = new FDistribution(df1, df2);
    try {/*from  w w w  .j  a  v a 2  s .  c  o m*/
        ci[0] = 1.0 - ((1.0 - reliability.value()) * fDist.inverseCumulativeProbability(0.975));
        ci[1] = 1.0 - ((1.0 - reliability.value()) * fDist.inverseCumulativeProbability(0.025));
    } catch (Exception ex) {
        ci[0] = Double.NaN;
        ci[1] = Double.NaN;
    }

    return ci;
}

From source file:com.rapidminer.operator.learner.functions.linear.IterativeTTestLinearRegressionMethod.java

@Override
public LinearRegressionResult applyMethod(LinearRegression regression, boolean useBias, double ridge,
        ExampleSet exampleSet, boolean[] isUsedAttribute, int numberOfExamples, int numberOfUsedAttributes,
        double[] means, double labelMean, double[] standardDeviations, double labelStandardDeviation,
        double[] coefficientsOnFullData, double errorOnFullData)
        throws UndefinedParameterError, ProcessStoppedException {
    int maxIterations = regression.getParameterAsInt(PARAMETER_MAX_ITERATIONS);
    double alphaForward = regression.getParameterAsDouble(PARAMETER_FORWARD_SELECTION_THRESHOLD);
    double alphaBackward = regression.getParameterAsDouble(PARAMETER_BACKWARD_SELECTION_THRESHOLD);

    FDistribution fdistribution;/*from w  w w. ja  v a 2 s .  c om*/
    // check if the F-distribution can be calculated
    int secondDegreeOfFreedom = exampleSet.size() - coefficientsOnFullData.length;
    if (secondDegreeOfFreedom > 0) {
        fdistribution = new FDistribution(1, secondDegreeOfFreedom);
    } else {
        fdistribution = null;
    }

    double generalCorrelation = regression.getCorrelation(exampleSet, isUsedAttribute, coefficientsOnFullData,
            useBias);
    generalCorrelation *= generalCorrelation;

    // building data structures
    boolean[] isAllowedToUse = isUsedAttribute;
    // initialize array for checking for change
    boolean[] isLastRoundUsed = new boolean[isUsedAttribute.length];
    boolean[] isToUseNextRound = new boolean[isUsedAttribute.length];
    isUsedAttribute = new boolean[isUsedAttribute.length];

    // do until nothing changes or max rounds exceeded
    int iteration = 0;
    while (iteration == 0
            || iteration < maxIterations && isSelectionDiffering(isUsedAttribute, isLastRoundUsed)) {
        System.arraycopy(isUsedAttribute, 0, isLastRoundUsed, 0, isUsedAttribute.length);

        // first do forward selection for all single non-selected and
        // allowed attributes
        int coefficientIndex = 0;
        for (int i = 0; i < isAllowedToUse.length; i++) {
            if (isAllowedToUse[i] && !isUsedAttribute[i]) {
                // check if this not selected one will receive significant coefficient
                isUsedAttribute[i] = true;
                double[] coefficients = regression.performRegression(exampleSet, isUsedAttribute, means,
                        labelMean, ridge, useBias);
                // only if it is possible to calculate the probabilities, the p-value for this
                // attribute is checked
                if (fdistribution != null) {
                    double pValue = getPValue(coefficients[coefficientIndex], i, regression, useBias, ridge,
                            exampleSet, isUsedAttribute, standardDeviations, labelStandardDeviation,
                            fdistribution, generalCorrelation);
                    if (1.0d - pValue <= alphaForward) {
                        isToUseNextRound[i] = true;
                    }
                }
                isUsedAttribute[i] = false;
            } else if (isUsedAttribute[i]) {
                coefficientIndex++;
            }
        }

        // now add all that we have remembered to use
        for (int i = 0; i < isUsedAttribute.length; i++) {
            isUsedAttribute[i] |= isToUseNextRound[i];
            isToUseNextRound[i] = false;
        }

        // now we have to deselect all that do not fulfill t-test in combination
        {
            double[] coefficients = regression.performRegression(exampleSet, isUsedAttribute, means, labelMean,
                    ridge, useBias);
            isUsedAttribute = filterByPValue(regression, useBias, ridge, exampleSet, isUsedAttribute, means,
                    labelMean, standardDeviations, labelStandardDeviation, coefficients,
                    alphaBackward).isUsedAttribute;
        }

        iteration++;
    }

    // calculate result
    LinearRegressionResult result = new LinearRegressionResult();
    result.isUsedAttribute = isUsedAttribute;
    result.coefficients = regression.performRegression(exampleSet, isUsedAttribute, means, labelMean, ridge,
            useBias);
    result.error = regression.getSquaredError(exampleSet, isUsedAttribute, result.coefficients, useBias);
    return result;
}

From source file:com.rapidminer.operator.learner.functions.linear.TTestLinearRegressionMethod.java

/**
 * This method filters the selected attributes depending on their p-value in respect to the
 * significance niveau alpha.//from  www  . jav a  2s  . c om
 *
 * @throws ProcessStoppedException
 */
protected LinearRegressionResult filterByPValue(LinearRegression regression, boolean useBias, double ridge,
        ExampleSet exampleSet, boolean[] isUsedAttribute, double[] means, double labelMean,
        double[] standardDeviations, double labelStandardDeviation, double[] coefficientsOnFullData,
        double alpha) throws UndefinedParameterError, ProcessStoppedException {

    FDistribution fdistribution;
    // check if the F-distribution can be calculated
    int secondDegreeOfFreedom = exampleSet.size() - coefficientsOnFullData.length;
    if (secondDegreeOfFreedom > 0) {
        fdistribution = new FDistribution(1, secondDegreeOfFreedom);
    } else {
        fdistribution = null;
    }

    double generalCorrelation = regression.getCorrelation(exampleSet, isUsedAttribute, coefficientsOnFullData,
            useBias);
    generalCorrelation *= generalCorrelation;

    int index = 0;
    for (int i = 0; i < isUsedAttribute.length; i++) {
        if (isUsedAttribute[i]) {
            double coefficient = coefficientsOnFullData[index];

            // only if it is possible to calculate the probabilities, the alpha value for this
            // attribute is checked
            if (fdistribution != null) {
                double probability = getPValue(coefficient, i, regression, useBias, ridge, exampleSet,
                        isUsedAttribute, standardDeviations, labelStandardDeviation, fdistribution,
                        generalCorrelation);
                if (1.0d - probability > alpha) {
                    isUsedAttribute[i] = false;
                }
                index++;
            } else {
                isUsedAttribute[i] = false;
            }
        }
    }
    LinearRegressionResult result = new LinearRegressionResult();
    result.isUsedAttribute = isUsedAttribute;
    result.coefficients = regression.performRegression(exampleSet, isUsedAttribute, means, labelMean, ridge,
            useBias);
    result.error = regression.getSquaredError(exampleSet, isUsedAttribute, result.coefficients, useBias);
    return result;
}

From source file:gedi.util.math.stat.testing.OneWayAnova.java

public double getPvalue() {
    FDistribution fdist = new FDistribution(getFactorDof(), getErrorDof());
    if (getFactorDof() < 1 || getErrorDof() < 1)
        return Double.NaN;
    return 1 - fdist.cumulativeProbability(getFStatistic());
}

From source file:com.rapidminer.operator.validation.significance.TTestSignificanceTestOperator.java

private double getProbability(PerformanceCriterion pc1, PerformanceCriterion pc2) {
    double totalDeviation = ((pc1.getAverageCount() - 1) * pc1.getVariance()
            + (pc2.getAverageCount() - 1) * pc2.getVariance())
            / (pc1.getAverageCount() + pc2.getAverageCount() - 2);
    double factor = 1.0d / (1.0d / pc1.getAverageCount() + 1.0d / pc2.getAverageCount());
    double diff = pc1.getAverage() - pc2.getAverage();
    double t = factor * diff * diff / totalDeviation;
    int secondDegreeOfFreedom = pc1.getAverageCount() + pc2.getAverageCount() - 2;
    double prob;/*from w w  w. j  a  va2  s.  co m*/
    // make sure the F-distribution is well defined
    if (secondDegreeOfFreedom > 0) {
        FDistribution fDist = new FDistribution(1, secondDegreeOfFreedom);
        prob = 1 - fDist.cumulativeProbability(t);
    } else {
        // in this case the probability cannot calculated correctly and a 1 is returned, as
        // this result is not significant
        prob = 1;
    }

    return prob;
}

From source file:edu.cudenver.bios.power.glmm.GLMMTest.java

/**
 * Fit the model/*from  www  .  j  a v  a  2 s.co m*/
 * @return model fit object
 */
public ModelFit getModelFit() {
    double fobs = getObservedF(GLMMTest.DistributionType.DATA_ANALYSIS_NULL);

    // get the p-value from a central F distribution
    double ndf = getNumeratorDF(GLMMTest.DistributionType.DATA_ANALYSIS_NULL);
    if (Double.isNaN(ndf)) {
        throw new IllegalArgumentException("numerator DF is NaN");
    }
    double ddf = getDenominatorDF(GLMMTest.DistributionType.DATA_ANALYSIS_NULL);
    if (Double.isNaN(ddf)) {
        throw new IllegalArgumentException("denominator DF is NaN");
    }

    FDistribution fdist = new FDistribution(ndf, ddf);
    double pvalue = 1 - fdist.cumulativeProbability(fobs);

    return new ModelFit(pvalue, fobs, ndf, ddf, sigmaError, beta);
}

From source file:edu.cudenver.bios.power.glmm.GLMMTest.java

/**
 * Calculate the critical F value under the specified distribution
 *
 * @param type distribution type//from w w w.  j av a  2  s. co m
 * @param alpha type I error level
 * @return critical F
 *
 */
public double getCriticalF(DistributionType type, double alpha) throws IllegalArgumentException {
    double ndf = getNumeratorDF(type);
    if (Double.isNaN(ndf)) {
        throw new IllegalArgumentException("numerator DF is NaN");
    }
    double ddf = getDenominatorDF(type);
    if (Double.isNaN(ddf)) {
        throw new IllegalArgumentException("denominator DF is NaN");
    }

    FDistribution centralFDist = new FDistribution(ndf, ddf);
    double fcrit = centralFDist.inverseCumulativeProbability(1 - alpha);
    return fcrit;
}

From source file:edu.stanford.cfuller.imageanalysistools.filter.VariableSizeMeanFilter.java

protected boolean shouldSubDivide(OcttreeNode node, Image im, Image laplacianFiltered) {

    im.setBoxOfInterest(node.getBoxMin(), node.getBoxMax());
    laplacianFiltered.setBoxOfInterest(node.getBoxMin(), node.getBoxMax());

    double l_sum = 0;
    double sum = 0;
    double count = 0;

    for (ImageCoordinate ic : im) {
        l_sum += laplacianFiltered.getValue(ic);
        sum += im.getValue(ic);//from   w  w w  . j  a v  a  2 s . c om
        count++;

    }

    if (count == 1)
        return false;

    l_sum /= count;
    sum /= count;

    double l_var = 0;
    double var = 0;

    for (ImageCoordinate ic : im) {

        l_var += Math.pow(laplacianFiltered.getValue(ic) - l_sum, 2);
        var += Math.pow(im.getValue(ic) - sum, 2);

    }

    l_var /= (count - 1);
    var /= (count - 1);

    im.clearBoxOfInterest();
    laplacianFiltered.clearBoxOfInterest();

    double cutoff = 0.0001;

    double smallerVar = var < l_var ? var : l_var;
    double largerVar = var > l_var ? var : l_var;
    try {

        FDistribution f = new FDistribution(count - 1, count - 1);
        double valueAtLowerCutoff = f.inverseCumulativeProbability(cutoff);
        double valueAtUpperCutoff = f.inverseCumulativeProbability(1 - cutoff);
        boolean result = (smallerVar / largerVar > valueAtUpperCutoff
                || smallerVar / largerVar < valueAtLowerCutoff);
        return result;

    } catch (MathIllegalArgumentException e) {
        LoggingUtilities.getLogger()
                .severe("Exception while calculating variable size mean QO partition: " + e.getMessage());
        e.printStackTrace();
        return false;
    }
}