Example usage for org.apache.commons.math.stat.inference OneWayAnova anovaPValue

Introduction

In this page you can find the example usage for org.apache.commons.math.stat.inference OneWayAnova anovaPValue.

Prototype

double anovaPValue(Collection<double[]> categoryData) throws IllegalArgumentException, MathException;

Source Link

Document

Computes the ANOVA P-value for a collection of double[] arrays.

Usage

From source file:rs.fon.whibo.GDT.component.removeInsignificantAttributes.FTestNumerical.java

public LinkedList<Attribute> removeAttributes(ExampleSet exampleSet,
        LinkedList<Attribute> attributesForSplitting) {
    // checks if the example set is pure, and if it is, it exits the method
    Attribute label = exampleSet.getAttributes().getLabel();
    if (Tools.getAllCategories(exampleSet, label).size() < 2)
        return attributesForSplitting;

    // selects the attributes to be evaluated for removal (by calculating
    // F-test probability for each attribute)
    ArrayList<Attribute> attributesToRemove = new ArrayList<Attribute>();
    ArrayList<Double> attributeProbabilities = new ArrayList<Double>();
    for (Attribute attr : attributesForSplitting)
        if (attr.isNumerical()) {
            // calculate F-test probability of the attribute
            double probability = 0;
            try {

                OneWayAnova fTest = new OneWayAnovaImpl();
                List<double[]> paramForFTest = getArraysByLabel(exampleSet, attr);

                // tests if no arrays for f-test has fewer that 2 elements
                boolean fTestImpossible = false;
                for (double[] i : paramForFTest)
                    if (i.length < 2)
                        fTestImpossible = true;

                // calculates ftest probability
                if (!fTestImpossible)
                    probability = fTest.anovaPValue(paramForFTest);

            } catch (Exception e) {
                // System.out.println("Error in calculating math formula (FTest)");
            }//  w  w w.  j  a v  a  2s.c  o  m
            // add the attribute to the list
            attributesToRemove.add(attr);
            attributeProbabilities.add(new Double(probability));
        }

    if (attributesToRemove.size() == 0)
        return attributesForSplitting;

    // calculates the percentile of the required percentage. Percentile
    // variable in code represents the percentage of attributes to be kept
    // (not removed)
    double percentile;
    DescriptiveStatistics stat = new DescriptiveStatistics();
    for (Double d : attributeProbabilities)
        stat.addValue(d.doubleValue());
    percentile = stat.getPercentile((1 - Percentage_Remove) * 100);

    // evaluates attributes and chooses the ones for removal (actually saves
    // the ones not for removal)
    Iterator<Attribute> iattr = attributesToRemove.iterator();
    Iterator<Double> iprob = attributeProbabilities.iterator();
    while (iattr.hasNext()) {
        iattr.next();
        Double prob = iprob.next();
        if (Use_Percentage_Instead == 0) {
            if (prob <= Alpha_Value) {
                iattr.remove();
                iprob.remove();
            }
        } else {
            if (prob <= percentile) {
                iattr.remove();
                iprob.remove();
            }
        }
    }

    // removes the attributes
    for (Attribute attr : attributesToRemove)
        attributesForSplitting.remove(attr);
    return attributesForSplitting;

}

---