<html> <head> <meta http-equiv="Content-Type" content="text/html; charset=US-ASCII"> <title>Noncentral Beta Distribution</title> <link rel="stylesheet" href="../../../../../../../../../doc/src/boostbook.css" type="text/css"> <meta name="generator" content="DocBook XSL Stylesheets V1.74.0"> <link rel="home" href="../../../../index.html" title="Math Toolkit"> <link rel="up" href="../dists.html" title="Distributions"> <link rel="prev" href="negative_binomial_dist.html" title="Negative Binomial Distribution"> <link rel="next" href="nc_chi_squared_dist.html" title="Noncentral Chi-Squared Distribution"> </head> <body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"> <table cellpadding="2" width="100%"><tr> <td valign="top"><img alt="Boost C++ Libraries" width="277" height="86" src="../../../../../../../../../boost.png"></td> <td align="center"><a href="../../../../../../../../../index.html">Home</a></td> <td align="center"><a href="../../../../../../../../../libs/libraries.htm">Libraries</a></td> <td align="center"><a href="http://www.boost.org/users/people.html">People</a></td> <td align="center"><a href="http://www.boost.org/users/faq.html">FAQ</a></td> <td align="center"><a href="../../../../../../../../../more/index.htm">More</a></td> </tr></table> <hr> <div class="spirit-nav"> <a accesskey="p" href="negative_binomial_dist.html"><img src="../../../../../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../dists.html"><img src="../../../../../../../../../doc/src/images/up.png" alt="Up"></a><a accesskey="h" href="../../../../index.html"><img src="../../../../../../../../../doc/src/images/home.png" alt="Home"></a><a accesskey="n" href="nc_chi_squared_dist.html"><img src="../../../../../../../../../doc/src/images/next.png" alt="Next"></a> </div> <div class="section" lang="en"> <div class="titlepage"><div><div><h5 class="title"> <a name="math_toolkit.dist.dist_ref.dists.nc_beta_dist"></a><a class="link" href="nc_beta_dist.html" title="Noncentral Beta Distribution"> Noncentral Beta Distribution</a> </h5></div></div></div> <p> </p> <pre class="programlisting"><span class="preprocessor">#include</span> <span class="special"><</span><span class="identifier">boost</span><span class="special">/</span><span class="identifier">math</span><span class="special">/</span><span class="identifier">distributions</span><span class="special">/</span><span class="identifier">non_central_beta</span><span class="special">.</span><span class="identifier">hpp</span><span class="special">></span></pre> <p> </p> <pre class="programlisting"><span class="keyword">namespace</span> <span class="identifier">boost</span><span class="special">{</span> <span class="keyword">namespace</span> <span class="identifier">math</span><span class="special">{</span> <span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">RealType</span> <span class="special">=</span> <span class="keyword">double</span><span class="special">,</span> <span class="keyword">class</span> <a class="link" href="../../../policy.html" title="Policies">Policy</a> <span class="special">=</span> <a class="link" href="../../../policy/pol_ref/pol_ref_ref.html" title="Policy Class Reference">policies::policy<></a> <span class="special">></span> <span class="keyword">class</span> <span class="identifier">non_central_beta_distribution</span><span class="special">;</span> <span class="keyword">typedef</span> <span class="identifier">non_central_beta_distribution</span><span class="special"><></span> <span class="identifier">non_central_beta</span><span class="special">;</span> <span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">RealType</span><span class="special">,</span> <span class="keyword">class</span> <a class="link" href="../../../policy.html" title="Policies">Policy</a><span class="special">></span> <span class="keyword">class</span> <span class="identifier">non_central_beta_distribution</span> <span class="special">{</span> <span class="keyword">public</span><span class="special">:</span> <span class="keyword">typedef</span> <span class="identifier">RealType</span> <span class="identifier">value_type</span><span class="special">;</span> <span class="keyword">typedef</span> <span class="identifier">Policy</span> <span class="identifier">policy_type</span><span class="special">;</span> <span class="comment">// Constructor: </span> <span class="identifier">non_central_beta_distribution</span><span class="special">(</span><span class="identifier">RealType</span> <span class="identifier">alpha</span><span class="special">,</span> <span class="identifier">RealType</span> <span class="identifier">beta</span><span class="special">,</span> <span class="identifier">RealType</span> <span class="identifier">lambda</span><span class="special">);</span> <span class="comment">// Accessor to shape parameters: </span> <span class="identifier">RealType</span> <span class="identifier">alpha</span><span class="special">()</span><span class="keyword">const</span><span class="special">;</span> <span class="identifier">RealType</span> <span class="identifier">beta</span><span class="special">()</span><span class="keyword">const</span><span class="special">;</span> <span class="comment">// Accessor to non-centrality parameter lambda: </span> <span class="identifier">RealType</span> <span class="identifier">non_centrality</span><span class="special">()</span><span class="keyword">const</span><span class="special">;</span> <span class="special">};</span> <span class="special">}}</span> <span class="comment">// namespaces </span></pre> <p> The noncentral beta distribution is a generalization of the <a class="link" href="beta_dist.html" title="Beta Distribution">Beta Distribution</a>. </p> <p> It is defined as the ratio X = χ<sub>m</sub><sup>2</sup>(λ) / (χ<sub>m</sub><sup>2</sup>(λ) + χ<sub>n</sub><sup>2</sup>) where χ<sub>m</sub><sup>2</sup>(λ) is a noncentral χ<sup>2</sup> random variable with <span class="emphasis"><em>m</em></span> degrees of freedom, and χ<sub>n</sub><sup>2</sup> is a central χ<sup>2</sup> random variable with <span class="emphasis"><em>n</em></span> degrees of freedom. </p> <p> This gives a PDF that can be expressed as a Poisson mixture of beta distribution PDFs: </p> <p> <span class="inlinemediaobject"><img src="../../../../../equations/nc_beta_ref1.png"></span> </p> <p> where P(i;λ/2) is the discrete Poisson probablity at <span class="emphasis"><em>i</em></span>, with mean λ/2, and I<sub>x</sub><sup>'</sup>(α, β) is the derivative of the incomplete beta function. This leads to the usual form of the CDF as: </p> <p> <span class="inlinemediaobject"><img src="../../../../../equations/nc_beta_ref2.png"></span> </p> <p> The following graph illustrates how the distribution changes for different values of λ: </p> <p> <span class="inlinemediaobject"><img src="../../../../../graphs/nc_beta_pdf.png" align="middle"></span> </p> <a name="math_toolkit.dist.dist_ref.dists.nc_beta_dist.member_functions"></a><h5> <a name="id1047231"></a> <a class="link" href="nc_beta_dist.html#math_toolkit.dist.dist_ref.dists.nc_beta_dist.member_functions">Member Functions</a> </h5> <pre class="programlisting"><span class="identifier">non_central_beta_distribution</span><span class="special">(</span><span class="identifier">RealType</span> <span class="identifier">a</span><span class="special">,</span> <span class="identifier">RealType</span> <span class="identifier">b</span><span class="special">,</span> <span class="identifier">RealType</span> <span class="identifier">lambda</span><span class="special">);</span> </pre> <p> Constructs a noncentral beta distribution with shape parameters <span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span> and non-centrality parameter <span class="emphasis"><em>lambda</em></span>. </p> <p> Requires a > 0, b > 0 and lambda >= 0, otherwise calls <a class="link" href="../../../main_overview/error_handling.html#domain_error">domain_error</a>. </p> <pre class="programlisting"><span class="identifier">RealType</span> <span class="identifier">alpha</span><span class="special">()</span><span class="keyword">const</span><span class="special">;</span> </pre> <p> Returns the parameter <span class="emphasis"><em>a</em></span> from which this object was constructed. </p> <pre class="programlisting"><span class="identifier">RealType</span> <span class="identifier">beta</span><span class="special">()</span><span class="keyword">const</span><span class="special">;</span> </pre> <p> Returns the parameter <span class="emphasis"><em>b</em></span> from which this object was constructed. </p> <pre class="programlisting"><span class="identifier">RealType</span> <span class="identifier">non_centrality</span><span class="special">()</span><span class="keyword">const</span><span class="special">;</span> </pre> <p> Returns the parameter <span class="emphasis"><em>lambda</em></span> from which this object was constructed. </p> <a name="math_toolkit.dist.dist_ref.dists.nc_beta_dist.non_member_accessors"></a><h5> <a name="id1047417"></a> <a class="link" href="nc_beta_dist.html#math_toolkit.dist.dist_ref.dists.nc_beta_dist.non_member_accessors">Non-member Accessors</a> </h5> <p> Most of the <a class="link" href="../nmp.html" title="Non-Member Properties">usual non-member accessor functions</a> are supported: <a class="link" href="../nmp.html#math.dist.cdf">Cumulative Distribution Function</a>, <a class="link" href="../nmp.html#math.dist.pdf">Probability Density Function</a>, <a class="link" href="../nmp.html#math.dist.quantile">Quantile</a>, <a class="link" href="../nmp.html#math.dist.median">median</a>, <a class="link" href="../nmp.html#math.dist.mode">mode</a>, <a class="link" href="../nmp.html#math.dist.hazard">Hazard Function</a>, <a class="link" href="../nmp.html#math.dist.chf">Cumulative Hazard Function</a>, <a class="link" href="../nmp.html#math.dist.range">range</a> and <a class="link" href="../nmp.html#math.dist.support">support</a>. </p> <p> However, the following are not currently implemented: <a class="link" href="../nmp.html#math.dist.mean">mean</a>, <a class="link" href="../nmp.html#math.dist.variance">variance</a>, <a class="link" href="../nmp.html#math.dist.sd">standard deviation</a>, <a class="link" href="../nmp.html#math.dist.skewness">skewness</a>, <a class="link" href="../nmp.html#math.dist.kurtosis">kurtosis</a> and <a class="link" href="../nmp.html#math.dist.kurtosis_excess">kurtosis_excess</a>. </p> <p> The domain of the random variable is [0, 1]. </p> <a name="math_toolkit.dist.dist_ref.dists.nc_beta_dist.accuracy"></a><h5> <a name="id1047518"></a> <a class="link" href="nc_beta_dist.html#math_toolkit.dist.dist_ref.dists.nc_beta_dist.accuracy">Accuracy</a> </h5> <p> The following table shows the peak errors (in units of <a href="http://en.wikipedia.org/wiki/Machine_epsilon" target="_top">epsilon</a>) found on various platforms with various floating point types. No comparison to the <a href="http://www.r-project.org/" target="_top">R-2.5.1 Math library</a>, or to the FORTRAN implementations of AS226 or AS310 are given since these appear to only guarantee absolute error: this would causes our test harness to assign an <span class="emphasis"><em>"infinite"</em></span> error to these libraries for some of our test values when measuring <span class="emphasis"><em>relative error</em></span>. Unless otherwise specified any floating-point type that is narrower than the one shown will have <a class="link" href="../../../backgrounders/relative_error.html#zero_error">effectively zero error</a>. </p> <div class="table"> <a name="id1047553"></a><p class="title"><b>Table 13. Errors In CDF of the Noncentral Beta</b></p> <div class="table-contents"><table class="table" summary="Errors In CDF of the Noncentral Beta"> <colgroup> <col> <col> <col> <col> </colgroup> <thead><tr> <th> <p> Significand Size </p> </th> <th> <p> Platform and Compiler </p> </th> <th> <p> α, β,λ < 200 </p> </th> <th> <p> α,β,λ > 200 </p> </th> </tr></thead> <tbody> <tr> <td> <p> 53 </p> </td> <td> <p> Win32, Visual C++ 8 </p> </td> <td> <p> Peak=620 Mean=22 </p> </td> <td> <p> Peak=8670 Mean=1040 </p> </td> </tr> <tr> <td> <p> 64 </p> </td> <td> <p> RedHat Linux IA32, gcc-4.1.1 </p> </td> <td> <p> Peak=825 Mean=50 </p> </td> <td> <p> Peak=2.5x10<sup>4</sup> Mean=4000 </p> </td> </tr> <tr> <td> <p> 64 </p> </td> <td> <p> Redhat Linux IA64, gcc-3.4.4 </p> </td> <td> <p> Peak=825 Mean=30 </p> </td> <td> <p> Peak=1.7x10<sup>4</sup> Mean=2500 </p> </td> </tr> <tr> <td> <p> 113 </p> </td> <td> <p> HPUX IA64, aCC A.06.06 </p> </td> <td> <p> Peak=420 Mean=50 </p> </td> <td> <p> Peak=9200 Mean=1200 </p> </td> </tr> </tbody> </table></div> </div> <br class="table-break"><p> Error rates for the PDF, the complement of the CDF and for the quantile functions are broadly similar. </p> <a name="math_toolkit.dist.dist_ref.dists.nc_beta_dist.tests"></a><h5> <a name="id1047759"></a> <a class="link" href="nc_beta_dist.html#math_toolkit.dist.dist_ref.dists.nc_beta_dist.tests">Tests</a> </h5> <p> There are two sets of test data used to verify this implementation: firstly we can compare with a few sample values generated by the <a href="http://www.r-project.org/" target="_top">R library</a>. Secondly, we have tables of test data, computed with this implementation and using interval arithmetic - this data should be accurate to at least 50 decimal digits - and is the used for our accuracy tests. </p> <a name="math_toolkit.dist.dist_ref.dists.nc_beta_dist.implementation"></a><h5> <a name="id1047783"></a> <a class="link" href="nc_beta_dist.html#math_toolkit.dist.dist_ref.dists.nc_beta_dist.implementation">Implementation</a> </h5> <p> The CDF and its complement are evaluated as follows: </p> <p> First we determine which of the two values (the CDF or its complement) is likely to be the smaller, the crossover point is taken to be the mean of the distribution: for this we use the approximation due to: R. Chattamvelli and R. Shanmugam, "Algorithm AS 310: Computing the Non-Central Beta Distribution Function", Applied Statistics, Vol. 46, No. 1. (1997), pp. 146-156. </p> <p> <span class="inlinemediaobject"><img src="../../../../../equations/nc_beta_ref3.png"></span> </p> <p> Then either the CDF or its complement is computed using the relations: </p> <p> <span class="inlinemediaobject"><img src="../../../../../equations/nc_beta_ref4.png"></span> </p> <p> The summation is performed by starting at i = λ/2, and then recursing in both directions, using the usual recurrence relations for the Poisson PDF and incomplete beta functions. This is the "Method 2" described by: </p> <p> Denise Benton and K. Krishnamoorthy, "Computing discrete mixtures of continuous distributions: noncentral chisquare, noncentral t and the distribution of the square of the sample multiple correlation coefficient", Computational Statistics & Data Analysis 43 (2003) 249-267. </p> <p> Specific applications of the above formulae to the noncentral beta distribution can be found in: </p> <p> Russell V. Lenth, "Algorithm AS 226: Computing Noncentral Beta Probabilities", Applied Statistics, Vol. 36, No. 2. (1987), pp. 241-244. </p> <p> H. Frick, "Algorithm AS R84: A Remark on Algorithm AS 226: Computing Non-Central Beta Probabilities", Applied Statistics, Vol. 39, No. 2. (1990), pp. 311-312. </p> <p> Ming Long Lam, "Remark AS R95: A Remark on Algorithm AS 226: Computing Non-Central Beta Probabilities", Applied Statistics, Vol. 44, No. 4. (1995), pp. 551-552. </p> <p> Harry O. Posten, "An Effective Algorithm for the Noncentral Beta Distribution Function", The American Statistician, Vol. 47, No. 2. (May, 1993), pp. 129-131. </p> <p> R. Chattamvelli, "A Note on the Noncentral Beta Distribution Function", The American Statistician, Vol. 49, No. 2. (May, 1995), pp. 231-234. </p> <p> Of these, the Posten reference provides the most complete overview, and includes the modification starting iteration at λ/2. </p> <p> The main difference between this implementation and the above references is the direct computation of the complement when most efficient to do so, and the accumulation of the sum to -1 rather than subtracting the result from 1 at the end: this can substantially reduce the number of iterations required when the result is near 1. </p> <p> The PDF is computed using the methodology of Benton and Krishnamoorthy and the relation: </p> <p> <span class="inlinemediaobject"><img src="../../../../../equations/nc_beta_ref1.png"></span> </p> <p> Quantiles are computed using a specially modified version of <a class="link" href="../../../toolkit/internals1/roots2.html" title="Root Finding Without Derivatives">bracket_and_solve_root</a>, starting the search for the root at the mean of the distribution. (A Cornish-Fisher type expansion was also tried, but while this gets quite close to the root in many cases, when it is wrong it tends to introduce quite pathological behaviour: more investigation in this area is probably warranted). </p> </div> <table xmlns:rev="http://www.cs.rpi.edu/~gregod/boost/tools/doc/revision" width="100%"><tr> <td align="left"></td> <td align="right"><div class="copyright-footer">Copyright © 2006 , 2007, 2008, 2009 John Maddock, Paul A. Bristow, Hubert Holin, Xiaogang Zhang, Bruno Lalande, Johan Råde, Gautam Sewani and Thijs van den Berg<p> Distributed under the Boost Software License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at <a href="http://www.boost.org/LICENSE_1_0.txt" target="_top">http://www.boost.org/LICENSE_1_0.txt</a>) </p> </div></td> </tr></table> <hr> <div class="spirit-nav"> <a accesskey="p" href="negative_binomial_dist.html"><img src="../../../../../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../dists.html"><img src="../../../../../../../../../doc/src/images/up.png" alt="Up"></a><a accesskey="h" href="../../../../index.html"><img src="../../../../../../../../../doc/src/images/home.png" alt="Home"></a><a accesskey="n" href="nc_chi_squared_dist.html"><img src="../../../../../../../../../doc/src/images/next.png" alt="Next"></a> </div> </body> </html>