3 <meta http-equiv="Content-Type" content="text/html; charset=US-ASCII">
4 <title>Generic operations common to all distributions are non-member functions</title>
5 <link rel="stylesheet" href="../../../math.css" type="text/css">
6 <meta name="generator" content="DocBook XSL Stylesheets V1.79.1">
7 <link rel="home" href="../../../index.html" title="Math Toolkit 2.11.0">
8 <link rel="up" href="../overview.html" title="Overview of Statistical Distributions">
9 <link rel="prev" href="objects.html" title="Distributions are Objects">
10 <link rel="next" href="complements.html" title="Complements are supported too - and when to use them">
12 <body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF">
13 <table cellpadding="2" width="100%"><tr>
14 <td valign="top"><img alt="Boost C++ Libraries" width="277" height="86" src="../../../../../../../boost.png"></td>
15 <td align="center"><a href="../../../../../../../index.html">Home</a></td>
16 <td align="center"><a href="../../../../../../../libs/libraries.htm">Libraries</a></td>
17 <td align="center"><a href="http://www.boost.org/users/people.html">People</a></td>
18 <td align="center"><a href="http://www.boost.org/users/faq.html">FAQ</a></td>
19 <td align="center"><a href="../../../../../../../more/index.htm">More</a></td>
22 <div class="spirit-nav">
23 <a accesskey="p" href="objects.html"><img src="../../../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../overview.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="complements.html"><img src="../../../../../../../doc/src/images/next.png" alt="Next"></a>
26 <div class="titlepage"><div><div><h4 class="title">
27 <a name="math_toolkit.stat_tut.overview.generic"></a><a class="link" href="generic.html" title="Generic operations common to all distributions are non-member functions">Generic operations
28 common to all distributions are non-member functions</a>
29 </h4></div></div></div>
31 Want to calculate the PDF (Probability Density Function) of a distribution?
34 <pre class="programlisting"><span class="identifier">pdf</span><span class="special">(</span><span class="identifier">my_dist</span><span class="special">,</span> <span class="identifier">x</span><span class="special">);</span> <span class="comment">// Returns PDF (density) at point x of distribution my_dist.</span>
37 Or how about the CDF (Cumulative Distribution Function):
39 <pre class="programlisting"><span class="identifier">cdf</span><span class="special">(</span><span class="identifier">my_dist</span><span class="special">,</span> <span class="identifier">x</span><span class="special">);</span> <span class="comment">// Returns CDF (integral from -infinity to point x)</span>
40 <span class="comment">// of distribution my_dist.</span>
43 And quantiles are just the same:
45 <pre class="programlisting"><span class="identifier">quantile</span><span class="special">(</span><span class="identifier">my_dist</span><span class="special">,</span> <span class="identifier">p</span><span class="special">);</span> <span class="comment">// Returns the value of the random variable x</span>
46 <span class="comment">// such that cdf(my_dist, x) == p.</span>
49 If you're wondering why these aren't member functions, it's to make the
50 library more easily extensible: if you want to add additional generic operations
51 - let's say the <span class="emphasis"><em>n'th moment</em></span> - then all you have to
52 do is add the appropriate non-member functions, overloaded for each implemented
55 <div class="tip"><table border="0" summary="Tip">
57 <td rowspan="2" align="center" valign="top" width="25"><img alt="[Tip]" src="../../../../../../../doc/src/images/tip.png"></td>
58 <th align="left">Tip</th>
60 <tr><td align="left" valign="top">
62 <span class="bold"><strong>Random numbers that approximate Quantiles of Distributions</strong></span>
65 If you want random numbers that are distributed in a specific way, for
66 example in a uniform, normal or triangular, see <a href="http://www.boost.org/libs/random/" target="_top">Boost.Random</a>.
69 Whilst in principal there's nothing to prevent you from using the quantile
70 function to convert a uniformly distributed random number to another
71 distribution, in practice there are much more efficient algorithms available
72 that are specific to random number generation.
77 For example, the binomial distribution has two parameters: n (the number
78 of trials) and p (the probability of success on any one trial).
81 The <code class="computeroutput"><span class="identifier">binomial_distribution</span></code>
82 constructor therefore has two parameters:
85 <code class="computeroutput"><span class="identifier">binomial_distribution</span><span class="special">(</span><span class="identifier">RealType</span> <span class="identifier">n</span><span class="special">,</span> <span class="identifier">RealType</span>
86 <span class="identifier">p</span><span class="special">);</span></code>
89 For this distribution the <a href="http://en.wikipedia.org/wiki/Random_variate" target="_top">random
90 variate</a> is k: the number of successes observed. The probability
91 density/mass function (pdf) is therefore written as <span class="emphasis"><em>f(k; n, p)</em></span>.
93 <div class="note"><table border="0" summary="Note">
95 <td rowspan="2" align="center" valign="top" width="25"><img alt="[Note]" src="../../../../../../../doc/src/images/note.png"></td>
96 <th align="left">Note</th>
98 <tr><td align="left" valign="top">
100 <span class="bold"><strong>Random Variates and Distribution Parameters</strong></span>
103 The concept of a <a href="http://en.wikipedia.org/wiki/Random_variable" target="_top">random
104 variable</a> is closely linked to the term <a href="http://en.wikipedia.org/wiki/Random_variate" target="_top">random
105 variate</a>: a random variate is a particular value (outcome) of
106 a random variable. and <a href="http://en.wikipedia.org/wiki/Parameter" target="_top">distribution
107 parameters</a> are conventionally distinguished (for example in Wikipedia
108 and Wolfram MathWorld) by placing a semi-colon or vertical bar) <span class="emphasis"><em>after</em></span>
109 the <a href="http://en.wikipedia.org/wiki/Random_variable" target="_top">random
110 variable</a> (whose value you 'choose'), to separate the variate
111 from the parameter(s) that defines the shape of the distribution.
114 For example, the binomial distribution probability distribution function
115 (PDF) is written as <span class="serif_italic"><span class="emphasis"><em>f(k| n, p)</em></span>
116 = Pr(K = k|n, p) = </span> probability of observing k successes out
117 of n trials. K is the <a href="http://en.wikipedia.org/wiki/Random_variable" target="_top">random
118 variable</a>, k is the <a href="http://en.wikipedia.org/wiki/Random_variate" target="_top">random
119 variate</a>, the parameters are n (trials) and p (probability).
123 <div class="note"><table border="0" summary="Note">
125 <td rowspan="2" align="center" valign="top" width="25"><img alt="[Note]" src="../../../../../../../doc/src/images/note.png"></td>
126 <th align="left">Note</th>
128 <tr><td align="left" valign="top"><p>
129 By convention, <a href="http://en.wikipedia.org/wiki/Random_variate" target="_top">random
130 variate</a> are lower case, usually k is integral, x if real, and
131 <a href="http://en.wikipedia.org/wiki/Random_variable" target="_top">random variable</a>
132 are upper case, K if integral, X if real. But this implementation treats
133 all as floating point values <code class="computeroutput"><span class="identifier">RealType</span></code>,
134 so if you really want an integral result, you must round: see note on
135 Discrete Probability Distributions below for details.
139 As noted above the non-member function <code class="computeroutput"><span class="identifier">pdf</span></code>
140 has one parameter for the distribution object, and a second for the random
141 variate. So taking our binomial distribution example, we would write:
144 <code class="computeroutput"><span class="identifier">pdf</span><span class="special">(</span><span class="identifier">binomial_distribution</span><span class="special"><</span><span class="identifier">RealType</span><span class="special">>(</span><span class="identifier">n</span><span class="special">,</span> <span class="identifier">p</span><span class="special">),</span> <span class="identifier">k</span><span class="special">);</span></code>
147 The ranges of <a href="http://en.wikipedia.org/wiki/Random_variate" target="_top">random
148 variate</a> values that are permitted and are supported can be tested
149 by using two functions <code class="computeroutput"><span class="identifier">range</span></code>
150 and <code class="computeroutput"><span class="identifier">support</span></code>.
153 The distribution (effectively the <a href="http://en.wikipedia.org/wiki/Random_variate" target="_top">random
154 variate</a>) is said to be 'supported' over a range that is <a href="http://en.wikipedia.org/wiki/Probability_distribution" target="_top">"the smallest
155 closed set whose complement has probability zero"</a>. MathWorld
156 uses the word 'defined' for this range. Non-mathematicians might say it
157 means the 'interesting' smallest range of random variate x that has the
158 cdf going from zero to unity. Outside are uninteresting zones where the
159 pdf is zero, and the cdf zero or unity.
162 For most distributions, with probability distribution functions one might
163 describe as 'well-behaved', we have decided that it is most useful for
164 the supported range to <span class="bold"><strong>exclude</strong></span> random
165 variate values like exact zero <span class="bold"><strong>if the end point is
166 discontinuous</strong></span>. For example, the Weibull (scale 1, shape 1) distribution
167 smoothly heads for unity as the random variate x declines towards zero.
168 But at x = zero, the value of the pdf is suddenly exactly zero, by definition.
169 If you are plotting the PDF, or otherwise calculating, zero is not the
170 most useful value for the lower limit of supported, as we discovered. So
171 for this, and similar distributions, we have decided it is most numerically
172 useful to use the closest value to zero, min_value, for the limit of the
173 supported range. (The <code class="computeroutput"><span class="identifier">range</span></code>
174 remains from zero, so you will still get <code class="computeroutput"><span class="identifier">pdf</span><span class="special">(</span><span class="identifier">weibull</span><span class="special">,</span> <span class="number">0</span><span class="special">)</span>
175 <span class="special">==</span> <span class="number">0</span></code>).
176 (Exponential and gamma distributions have similarly discontinuous functions).
179 Mathematically, the functions may make sense with an (+ or -) infinite
180 value, but except for a few special cases (in the Normal and Cauchy distributions)
181 this implementation limits random variates to finite values from the <code class="computeroutput"><span class="identifier">max</span></code> to <code class="computeroutput"><span class="identifier">min</span></code>
182 for the <code class="computeroutput"><span class="identifier">RealType</span></code>. (See
183 <a class="link" href="../../sf_implementation.html#math_toolkit.sf_implementation.handling_of_floating_point_infin">Handling
184 of Floating-Point Infinity</a> for rationale).
186 <div class="note"><table border="0" summary="Note">
188 <td rowspan="2" align="center" valign="top" width="25"><img alt="[Note]" src="../../../../../../../doc/src/images/note.png"></td>
189 <th align="left">Note</th>
191 <tr><td align="left" valign="top">
193 <span class="bold"><strong>Discrete Probability Distributions</strong></span>
196 Note that the <a href="http://en.wikipedia.org/wiki/Discrete_probability_distribution" target="_top">discrete
197 distributions</a>, including the binomial, negative binomial, Poisson
198 & Bernoulli, are all mathematically defined as discrete functions:
199 that is to say the functions <code class="computeroutput"><span class="identifier">cdf</span></code>
200 and <code class="computeroutput"><span class="identifier">pdf</span></code> are only defined
201 for integral values of the random variate.
204 However, because the method of calculation often uses continuous functions
205 it is convenient to treat them as if they were continuous functions,
206 and permit non-integral values of their parameters.
209 Users wanting to enforce a strict mathematical model may use <code class="computeroutput"><span class="identifier">floor</span></code> or <code class="computeroutput"><span class="identifier">ceil</span></code>
210 functions on the random variate prior to calling the distribution function.
213 The quantile functions for these distributions are hard to specify in
214 a manner that will satisfy everyone all of the time. The default behaviour
215 is to return an integer result, that has been rounded <span class="emphasis"><em>outwards</em></span>:
216 that is to say, lower quantiles - where the probablity is less than 0.5
217 are rounded down, while upper quantiles - where the probability is greater
218 than 0.5 - are rounded up. This behaviour ensures that if an X% quantile
219 is requested, then <span class="emphasis"><em>at least</em></span> the requested coverage
220 will be present in the central region, and <span class="emphasis"><em>no more than</em></span>
221 the requested coverage will be present in the tails.
224 This behaviour can be changed so that the quantile functions are rounded
225 differently, or return a real-valued result using <a class="link" href="../../pol_overview.html" title="Policy Overview">Policies</a>.
226 It is strongly recommended that you read the tutorial <a class="link" href="../../pol_tutorial/understand_dis_quant.html" title="Understanding Quantiles of Discrete Distributions">Understanding
227 Quantiles of Discrete Distributions</a> before using the quantile
228 function on a discrete distribtion. The <a class="link" href="../../pol_ref/discrete_quant_ref.html" title="Discrete Quantile Policies">reference
229 docs</a> describe how to change the rounding policy for these distributions.
232 For similar reasons continuous distributions with parameters like "degrees
233 of freedom" that might appear to be integral, are treated as real
234 values (and are promoted from integer to floating-point if necessary).
235 In this case however, there are a small number of situations where non-integral
236 degrees of freedom do have a genuine meaning.
241 <table xmlns:rev="http://www.cs.rpi.edu/~gregod/boost/tools/doc/revision" width="100%"><tr>
242 <td align="left"></td>
243 <td align="right"><div class="copyright-footer">Copyright © 2006-2019 Nikhar
244 Agrawal, Anton Bikineev, Paul A. Bristow, Marco Guazzone, Christopher Kormanyos,
245 Hubert Holin, Bruno Lalande, John Maddock, Jeremy Murphy, Matthew Pulver, Johan
246 Råde, Gautam Sewani, Benjamin Sobotta, Nicholas Thompson, Thijs van den Berg,
247 Daryle Walker and Xiaogang Zhang<p>
248 Distributed under the Boost Software License, Version 1.0. (See accompanying
249 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>)
254 <div class="spirit-nav">
255 <a accesskey="p" href="objects.html"><img src="../../../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../overview.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="complements.html"><img src="../../../../../../../doc/src/images/next.png" alt="Next"></a>