// <copyright file="StudentT.cs" company="Math.NET">
|
// Math.NET Numerics, part of the Math.NET Project
|
// http://numerics.mathdotnet.com
|
// http://github.com/mathnet/mathnet-numerics
|
//
|
// Copyright (c) 2009-2020 Math.NET
|
//
|
// Permission is hereby granted, free of charge, to any person
|
// obtaining a copy of this software and associated documentation
|
// files (the "Software"), to deal in the Software without
|
// restriction, including without limitation the rights to use,
|
// copy, modify, merge, publish, distribute, sublicense, and/or sell
|
// copies of the Software, and to permit persons to whom the
|
// Software is furnished to do so, subject to the following
|
// conditions:
|
//
|
// The above copyright notice and this permission notice shall be
|
// included in all copies or substantial portions of the Software.
|
//
|
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
|
// OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
|
// HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
|
// WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
|
// OTHER DEALINGS IN THE SOFTWARE.
|
// </copyright>
|
|
using IStation.Numerics.Random;
|
using System;
|
using System.Collections.Generic;
|
|
namespace IStation.Numerics.Distributions
|
{
|
/// <summary>
|
/// Continuous Univariate Skewed Generalized Error Distribution (SGED).
|
/// Implements the univariate SSkewed Generalized Error Distribution. For details about this
|
/// distribution, see
|
/// <a href="https://en.wikipedia.org/wiki/Generalized_normal_distribution">
|
/// Wikipedia - Generalized Error Distribution</a>.
|
/// It includes Laplace, Normal and Student-t distributions.
|
/// This is the <see cref="SkewedGeneralizedT"/> distribution with q=Inf.
|
/// </summary>
|
/// <remarks><para>This implementation is based on the R package dsgt and corresponding viginette, see
|
/// <a href="">https://cran.r-project.org/web/packages/sgt/vignettes/sgt.pdf</a>. Compared to that
|
/// implementation, the options for mean adjustment and variance adjustment are always true.
|
/// The location (μ) is the mean of the distribution.
|
/// The scale (σ) squared is the variance of the distribution.
|
/// </para>
|
/// <para>The distribution will use the <see cref="System.Random"/> by
|
/// default. Users can get/set the random number generator by using the
|
/// <see cref="RandomSource"/> property.</para>
|
/// <para>The statistics classes will check all the incoming parameters
|
/// whether they are in the allowed range.</para></remarks>
|
public class SkewedGeneralizedError : IContinuousDistribution
|
{
|
System.Random _random;
|
|
readonly double _skewness;
|
|
/// <summary>
|
/// Initializes a new instance of the SkewedGeneralizedError class. This is a generalized error distribution
|
/// with location=0.0, scale=1.0, skew=0.0 and p=2.0 (a standard normal distribution).
|
/// </summary>
|
public SkewedGeneralizedError()
|
{
|
_random = SystemRandomSource.Default;
|
Location = 0.0;
|
Scale = 1.0;
|
Skew = 0.0;
|
P = 2.0;
|
}
|
|
/// <summary>
|
/// Initializes a new instance of the SkewedGeneralizedT class with a particular location, scale, skew
|
/// and kurtosis parameters. Different parameterizations result in different distributions.
|
/// </summary>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
public SkewedGeneralizedError(double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
_random = SystemRandomSource.Default;
|
Location = location;
|
Scale = scale;
|
Skew = skew;
|
P = p;
|
|
_skewness = CalculateSkewness();
|
}
|
|
/// <summary>
|
/// Gets or sets the random number generator which is used to draw random samples.
|
/// </summary>
|
public System.Random RandomSource
|
{
|
get => _random;
|
set => _random = value ?? SystemRandomSource.Default;
|
}
|
|
/// <summary>
|
/// A string representation of the distribution.
|
/// </summary>
|
/// <returns>a string representation of the distribution.</returns>
|
public override string ToString()
|
{
|
return $"SkewedGeneralizedError(μ = {Location}, σ = {Scale}, λ = { Skew }, p = {P}";
|
}
|
|
/// <summary>
|
/// Tests whether the provided values are valid parameters for this distribution.
|
/// </summary>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
public static bool IsValidParameterSet(double location, double scale, double skew, double p)
|
{
|
return scale > 0.0 && skew > -1.0 && skew < 1.0 && p > 0.0 && !double.IsNaN(location);
|
}
|
|
/// <summary>
|
/// Gets the location (μ) of the Skewed Generalized t-distribution.
|
/// </summary>
|
public double Location { get; private set; }
|
|
/// <summary>
|
/// Gets the scale (σ) of the Skewed Generalized t-distribution. Range: σ > 0.
|
/// </summary>
|
public double Scale { get; private set; }
|
|
/// <summary>
|
/// Gets the skew (λ) of the Skewed Generalized t-distribution. Range: 1 > λ > -1.
|
/// </summary>
|
public double Skew { get; private set; }
|
|
/// <summary>
|
/// Gets the parameter that controls the kurtosis of the distribution. Range: p > 0.
|
/// </summary>
|
public double P { get; private set; }
|
|
// No skew implies Median=Mode=Mean
|
public double Mode =>
|
Skew == 0 ? Mean : Mean - AdjustAddend(AdjustScale(Scale, Skew, P), Skew, P);
|
|
public double Minimum => double.NegativeInfinity;
|
|
public double Maximum => double.PositiveInfinity;
|
|
// Mean=Location due to our adjustments made
|
public double Mean => Location;
|
|
// Variance=Scale*Scale due to our adjustments made
|
public double Variance => Scale * Scale;
|
|
public double StdDev => Scale;
|
|
public double Entropy => throw new NotImplementedException();
|
|
public double Skewness => _skewness;
|
|
// No skew implies Median=Mode=Mean
|
// Else find it via the point where CDF gives 0.5
|
public double Median =>
|
Skew == 0 ? Mean : InverseCumulativeDistribution(0.5);
|
|
double CalculateSkewness()
|
{
|
if (Skew == 0)
|
{
|
return 0.0;
|
}
|
|
var piPow = Math.Pow(Constants.Pi, 3.0 / 2.0);
|
var g1 = SpecialFunctions.Gamma(1.0 / P);
|
var g2 = SpecialFunctions.Gamma(0.5 + 1.0 / P);
|
var g3 = SpecialFunctions.Gamma(3.0 / P);
|
var g4 = SpecialFunctions.Gamma(4.0 / P);
|
|
var t1 = Skew * Scale * Scale * Scale / (piPow * g1);
|
var t2 = Math.Pow(2.0, (6.0 + P) / P) * Skew * Skew * Math.Pow(g2, 3.0) * g1;
|
var t3 = 3.0 * Math.Pow(4.0, 1.0 / P) * Constants.Pi * (1.0 + 3.0 * Skew * Skew) * g2 * g3;
|
var t4 = 4.0 * piPow * (1.0 + Skew * Skew) * g4;
|
|
return t1 * (t2 - t3 + t4);
|
}
|
|
static double AdjustScale(double scale, double skew, double p)
|
{
|
var g1 = SpecialFunctions.Gamma(3.0 / p);
|
var g2 = SpecialFunctions.Gamma(0.5 + 1.0 / p);
|
var g3 = SpecialFunctions.Gamma(1.0 / p);
|
var g4 = SpecialFunctions.Gamma(1.0 / p);
|
var n1 = Constants.Pi * (1.0 + 3.0 * skew * skew) * g1;
|
var n2 = Math.Pow(16.0, 1.0 / p) * skew * skew * Math.Pow(g2, 2) * g3;
|
var d = Constants.Pi * g4;
|
return scale / Math.Sqrt((n1 - n2) / d);
|
}
|
|
static double AdjustX(double x, double scale, double skew, double p)
|
{
|
return x + AdjustAddend(scale, skew, p);
|
}
|
|
static double AdjustAddend(double scale, double skew, double p)
|
{
|
return (Math.Pow(2.0, 2.0 / p) * scale * skew * SpecialFunctions.Gamma(1.0 / 2.0 + 1.0 / p)) /
|
Math.Sqrt(Constants.Pi);
|
}
|
|
public static double PDF(double location, double scale, double skew, double p, double x)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
scale = AdjustScale(scale, skew, p);
|
x = AdjustX(x, scale, skew, p);
|
|
// p / (2 * sigma * gamma(1 / p) * exp((abs(x - mu) / (sigma * (1 + lambda * sgn(x - mu)))) ^ p))
|
var d1 = Math.Abs(x - location);
|
var d2 = scale * (1.0 + skew * Math.Sign(x - location));
|
var d3 = 2.0 * scale * SpecialFunctions.Gamma(1.0 / p);
|
return p / (Math.Exp(Math.Pow(d1 / d2, p)) * d3);
|
}
|
|
public static double PDFLn(double location, double scale, double skew, double p, double x)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
scale = AdjustScale(scale, skew, p);
|
x = AdjustX(x, scale, skew, p);
|
|
return Math.Log(p) - Math.Log(2.0) - Math.Log(scale) - SpecialFunctions.GammaLn(1.0 / p) -
|
Math.Pow(Math.Abs(x - location) / (scale * (1.0 + skew * Math.Sign(x - location))), p);
|
}
|
|
public static double CDF(double location, double scale, double skew, double p, double x)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
scale = AdjustScale(scale, skew, p);
|
x = AdjustX(x, scale, skew, p) - location;
|
|
var flip = x < 0;
|
if (flip)
|
{
|
skew = -skew;
|
x = -x;
|
}
|
|
var res = (1.0 - skew) / 2.0 + (1.0 + skew) / 2.0 * Gamma.CDF(1.0 / p, 1.0, Math.Pow(x / (scale * (1.0 + skew)), p));
|
return flip ? 1.0 - res : res;
|
}
|
|
public static double InvCDF(double location, double scale, double skew, double p, double pr)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
scale = AdjustScale(scale, skew, p);
|
|
var flip = pr < (1.0 - skew) / 2.0;
|
var lambda = skew;
|
if (flip)
|
{
|
pr = 1.0 - pr;
|
lambda = -lambda;
|
}
|
|
var res = scale * (1.0 + lambda) * Math.Pow(Gamma.InvCDF(1.0 / p, 1.0, 2 * pr / (1.0 + lambda) + (lambda - 1.0) / (lambda + 1.0)), 1.0 / p);
|
|
if (flip)
|
res = -res;
|
res += location;
|
return res - AdjustAddend(scale, skew, p);
|
}
|
|
public double InverseCumulativeDistribution(double p)
|
{
|
return InvCDF(Location, Scale, Skew, P, p);
|
}
|
|
public double CumulativeDistribution(double x)
|
{
|
return CDF(Location, Scale, Skew, P, x);
|
}
|
|
public double Density(double x)
|
{
|
return PDF(Location, Scale, Skew, P, x);
|
}
|
|
public double DensityLn(double x)
|
{
|
return PDFLn(Location, Scale, Skew, P, x);
|
}
|
|
public double Sample()
|
{
|
return SampleUnchecked(_random, Location, Scale, Skew, P);
|
}
|
|
public void Samples(double[] values)
|
{
|
SamplesUnchecked(_random, values, Location, Scale, Skew, P);
|
}
|
|
public IEnumerable<double> Samples()
|
{
|
return SamplesUnchecked(_random, Location, Scale, Skew, P);
|
}
|
|
static double SampleUnchecked(System.Random rnd, double location, double scale, double skew, double p)
|
{
|
var u = ContinuousUniform.Sample(rnd, 0, 1);
|
return InvCDF(location, scale, skew, p, u);
|
}
|
|
static void SamplesUnchecked(System.Random rnd, double[] values, double location, double scale, double skew, double p)
|
{
|
for (int i = 0; i < values.Length; i++)
|
{
|
values[i] = SampleUnchecked(rnd, location, scale, skew, p);
|
}
|
}
|
|
static IEnumerable<double> SamplesUnchecked(System.Random rnd, double location, double scale, double skew, double p)
|
{
|
while (true)
|
{
|
yield return SampleUnchecked(rnd, location, scale, skew, p);
|
}
|
}
|
|
/// <summary>
|
/// Generates a sample from the Skew Generalized Error distribution.
|
/// </summary>
|
/// <param name="rnd">The random number generator to use.</param>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
/// <returns>a sample from the distribution.</returns>
|
public static double Sample(System.Random rnd, double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
return SampleUnchecked(rnd, location, scale, skew, p);
|
}
|
|
/// <summary>
|
/// Generates a sequence of samples from the Skew Generalized Error distribution using inverse transform.
|
/// </summary>
|
/// <param name="rnd">The random number generator to use.</param>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
/// <returns>a sequence of samples from the distribution.</returns>
|
public static IEnumerable<double> Samples(System.Random rnd, double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
return SamplesUnchecked(rnd, location, scale, skew, p);
|
}
|
|
/// <summary>
|
/// Fills an array with samples from the Skew Generalized Error distribution using inverse transform.
|
/// </summary>
|
/// <param name="rnd">The random number generator to use.</param>
|
/// <param name="values">The array to fill with the samples.</param>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
/// <returns>a sequence of samples from the distribution.</returns>
|
public static void Samples(System.Random rnd, double[] values, double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
SamplesUnchecked(rnd, values, location, scale, skew, p);
|
}
|
|
/// <summary>
|
/// Generates a sample from the Skew Generalized Error distribution.
|
/// </summary>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
/// <returns>a sample from the distribution.</returns>
|
public static double Sample(double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
return SampleUnchecked(SystemRandomSource.Default, location, scale, skew, p);
|
}
|
|
/// <summary>
|
/// Generates a sequence of samples from the Skew Generalized Error distribution using inverse transform.
|
/// </summary>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
/// <returns>a sequence of samples from the distribution.</returns>
|
public static IEnumerable<double> Samples(double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
return SamplesUnchecked(SystemRandomSource.Default, location, scale, skew, p);
|
}
|
|
/// <summary>
|
/// Fills an array with samples from the Skew Generalized Error distribution using inverse transform.
|
/// </summary>
|
/// <param name="values">The array to fill with the samples.</param>
|
/// <param name="location">The location (μ) of the distribution.</param>
|
/// <param name="scale">The scale (σ) of the distribution. Range: σ > 0.</param>
|
/// <param name="skew">The skew, 1 > λ > -1</param>
|
/// <param name="p">Parameter that controls kurtosis. Range: p > 0</param>
|
/// <returns>a sequence of samples from the distribution.</returns>
|
public static void Samples(double[] values, double location, double scale, double skew, double p)
|
{
|
if (!IsValidParameterSet(location, scale, skew, p))
|
{
|
throw new ArgumentException("Invalid parametrization for the distribution.");
|
}
|
|
SamplesUnchecked(SystemRandomSource.Default, values, location, scale, skew, p);
|
}
|
}
|
}
|