namespace Algorithms.Other;
/// Implementation of Welford's variance algorithm.
///
public class WelfordsVariance
{
///
/// Mean accumulates the mean of the entire dataset,
/// m2 aggregates the squared distance from the mean,
/// count aggregates the number of samples seen so far.
///
private int count;
public double Count => count;
private double mean;
public double Mean => count > 1 ? mean : double.NaN;
private double m2;
public double Variance => count > 1 ? m2 / count : double.NaN;
public double SampleVariance => count > 1 ? m2 / (count - 1) : double.NaN;
public WelfordsVariance()
{
count = 0;
mean = 0;
}
public WelfordsVariance(double[] values)
{
count = 0;
mean = 0;
AddRange(values);
}
public void AddValue(double newValue)
{
count++;
AddValueToDataset(newValue);
}
public void AddRange(double[] values)
{
var length = values.Length;
for (var i = 1; i <= length; i++)
{
count++;
AddValueToDataset(values[i - 1]);
}
}
private void AddValueToDataset(double newValue)
{
var delta1 = newValue - mean;
var newMean = mean + delta1 / count;
var delta2 = newValue - newMean;
m2 += delta1 * delta2;
mean = newMean;
}
}