RunningStatistics 3.0.0

RunningStatistics

Online (single pass) algorithms for statistical measures based on the Julia package OnlineStats.jl by Josh Day. Useful for streaming and big data.

This package is available on NuGet

Supports .NET Standard 2.0

List of Statistics

List of Distributions

Common Interface

All running statistics implement the following interfaces:

public interface IRunningStatistic
{
    public long Nobs { get; }

    public void Reset();

    public IRunningStatistic CloneEmpty();

    public IRunningStatistic Clone();

    public void UnsafeMerge(IRunningStatistic other);
}

public interface IRunningStatistic<TObs> : IRunningStatistic
{
    public void Fit(TObs value);

    public void Fit(TObs value, long count);

    public void Fit(IEnumerable<TObs> values);

    public void Fit(IEnumerable<KeyValuePair<TObs, long>> keyValuePairs);

    public new IRunningStatistic<TObs> CloneEmpty();

    public new IRunningStatistic<TObs> Clone();

    public void UnsafeMerge(IRunningStatistic<TObs> other);
}

public interface IRunningStatistic<TObs, TSelf> : IRunningStatistic<TObs>
    where TSelf : IRunningStatistic<TObs, TSelf>
{
    public new TSelf CloneEmpty();

    public new TSelf Clone();

    public void Merge(TSelf other);
}

The interface is layered as three interfaces of increasing specificity. IRunningStatistic is the most generic and allows for mixed running statistics to be in a collection. IRunningStatistic<TObs> expands by adding information about the type of observations that the running statistic can fit. Finally IRunningStatistic<TObs, TSelf> adds information about the concrete type of the running statistic, and allows for precise cloning and merging.

Abstract Base Class

The abstract class RunningStatisticBase implements the IRunningStatistic<TObs, TSelf> interface and provides default implementations when possible.

Examples

See the RunningStatistics.Examples project for more examples.

Basic Usage

using RunningStatistics;

var mean1 = new Mean();
var mean2 = new Mean();
var ecdf = new EmpiricalCdf();

var rng = new Random();
for (var i = 0; i < 1000; i++)
{
    var x = rng.NextDouble();
    
    mean1.Fit(x);
    mean2.Fit(2*x);
    
    ecdf.Fit(x);
}

mean1.Merge(mean2);
var q1 = ecdf.Quantile(0.25);

Upgrading from v2 to v3

If you are inheriting from RunningStatisticBase, the biggest change should be the removal of the GetStatsString method. Now you should just override the ToString method.

Before:

protected override string GetStatsString() => $"μ={Mean}, σ²={Variance}";

After:

public override string ToString() => base.ToString() + $" | μ={Mean}, σ²={Variance}";