com.espertech.esper.epl.agg.aggregator.AggregatorStddev Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of esper Show documentation

Complex event processing and event series analysis component

There is a newer version: 7.1.0

/**************************************************************************************
 * Copyright (C) 2006-2015 EsperTech Inc. All rights reserved.                        *
 * http://www.espertech.com/esper                                                          *
 * http://www.espertech.com                                                           *
 * ---------------------------------------------------------------------------------- *
 * The software in this package is published under the terms of the GPL license       *
 * a copy of which has been included with this distribution in the license.txt file.  *
 **************************************************************************************/
package com.espertech.esper.epl.agg.aggregator;

/**
 * Standard deviation always generates double-typed numbers.
 */
public class AggregatorStddev implements AggregationMethod
{
    protected double mean;
    protected double qn;
    protected long numDataPoints;

    public void clear()
    {
        mean = 0;
        numDataPoints = 0;
        qn = 0;
    }

    public void enter(Object object)
    {
        if (object == null)
        {
            return;
        }

        double p = ((Number) object).doubleValue();

        // compute running variance per Knuth's method
        if (numDataPoints == 0) {
            mean = p;
            qn = 0;
            numDataPoints = 1;
        }
        else {
            numDataPoints++;
            double oldmean = mean;
            mean += (p - mean)/numDataPoints;
            qn += (p - oldmean)*(p - mean);
        }
    }

    public void leave(Object object)
    {
        if (object == null)
        {
            return;
        }

        double p = ((Number) object).doubleValue();

        // compute running variance per Knuth's method
        if (numDataPoints <= 1) {
            clear();
        }
        else {
            numDataPoints--;
            double oldmean = mean;
            mean -= (p - mean)/numDataPoints;
            qn -= (p - oldmean)*(p - mean);
        }
    }

    public Object getValue()
    {
        if (numDataPoints < 2) {
            return null;
        }
        return Math.sqrt(qn / (numDataPoints - 1));
    }

}