iamchrisb / meanvar

Tiny Java utility to incrementally calculate Mean and Standard Deviation with a numerically stable algorithm.

tools4j-meanvar

Tiny Java utility to incrementally calculate Mean and Standard Deviation with a numerically stable algorithm. Contains a simple utility class to incrementally calculate moving average and moving standard deviation of a data series.

The implementation is based on Welford’s Algorithm given in Knuth Vol 2, p 232.

Examples

MeanVarianceSampler

final MeanVarianceSampler sampler = new MeanVarianceSampler();

double mean, var, stdDev;

sampler.add(1);
sampler.add(2.5);
sampler.add(3.22);
sampler.add(-6.72);
mean = sampler.getMean();
var = sampler.getVariance();
stdDev = sampler.getStdDev();

sampler.remove(2.5);
mean = sampler.getMean();
var = sampler.getVariance();
stdDev = sampler.getStdDev();

sampler.replace(3.22, 4.22);
mean = sampler.getMean();
var = sampler.getVariance();
stdDev = sampler.getStdDev();

MeanVarianceSlidingWindow

final int windowSize = 3;
final MeanVarianceSlidingWindow win = new MeanVarianceSlidingWindow(windowSize);

double mean, var, stdDev;

win.update(1);
win.update(2);
win.update(3);
mean = win.getMean();
var = win.getVariance();
stdDev = win.getStdDev();

//1 drops out now
win.update(4);
mean = win.getMean();
var = win.getVariance();
stdDev = win.getStdDev();

//2 drops out now
win.update(5);
mean = win.getMean();
var = win.getVariance();
stdDev = win.getStdDev();

More Information

• MIT License
• MeanVarianceSampler.java: Utility to add, remove or replace values in a running calculation of mean and variance
• MeanVarianceSlidingWindow.java: A fixed length sliding window to calculate moving average and moving standard deviation of a data series