WelfordsVariance.cs

namespaceAlgorithms.Other;

/// <summary>Implementation of Welford's variance algorithm.
/// </summary>
publicclassWelfordsVariance
{
/// <summary>
/// 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.
/// </summary>
privateintcount;

publicdoubleCount=>count;

privatedoublemean;

publicdoubleMean=>count>1?mean:double.NaN;

privatedoublem2;

publicdoubleVariance=>count>1?m2/count:double.NaN;

publicdoubleSampleVariance=>count>1?m2/(count-1):double.NaN;

publicWelfordsVariance()
{
count=0;
mean=0;
}

publicWelfordsVariance(double[]values)
{
count=0;
mean=0;
AddRange(values);
}

publicvoidAddValue(doublenewValue)
{
count++;
AddValueToDataset(newValue);
}

publicvoidAddRange(double[]values)
{
varlength=values.Length;
for(vari=1;i<=length;i++)
{
count++;
AddValueToDataset(values[i-1]);
}
}

privatevoidAddValueToDataset(doublenewValue)
{
vardelta1=newValue-mean;
varnewMean=mean+delta1/count;

vardelta2=newValue-newMean;
m2+=delta1*delta2;

mean=newMean;
}
}