| 1 |
// Copyright John Maddock 2006. |
| 2 |
// Use, modification and distribution are subject to the |
| 3 |
// Boost Software License, Version 1.0. (See accompanying file |
| 4 |
// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) |
| 5 |
|
| 6 |
// distributions.hpp provides definitions of the concept of a distribution |
| 7 |
// and non-member accessor functions that must be implemented by all distributions. |
| 8 |
// This is used to verify that |
| 9 |
// all the features of a distributions have been fully implemented. |
| 10 |
|
| 11 |
#ifndef BOOST_MATH_DISTRIBUTION_CONCEPT_HPP |
| 12 |
#define BOOST_MATH_DISTRIBUTION_CONCEPT_HPP |
| 13 |
|
| 14 |
#ifndef BOOST_MATH_STANDALONE |
| 15 |
|
| 16 |
#include <boost/math/distributions/complement.hpp> |
| 17 |
#include <boost/math/distributions/fwd.hpp> |
| 18 |
#ifdef _MSC_VER |
| 19 |
#pragma warning(push) |
| 20 |
#pragma warning(disable: 4100) |
| 21 |
#pragma warning(disable: 4510) |
| 22 |
#pragma warning(disable: 4610) |
| 23 |
#pragma warning(disable: 4189) // local variable is initialized but not referenced. |
| 24 |
#endif |
| 25 |
#include <boost/concept_check.hpp> |
| 26 |
#ifdef _MSC_VER |
| 27 |
#pragma warning(pop) |
| 28 |
#endif |
| 29 |
#include <utility> |
| 30 |
|
| 31 |
namespace boost{ |
| 32 |
namespace math{ |
| 33 |
|
| 34 |
namespace concepts |
| 35 |
{ |
| 36 |
// Begin by defining a concept archetype |
| 37 |
// for a distribution class: |
| 38 |
// |
| 39 |
template <class RealType> |
| 40 |
class distribution_archetype |
| 41 |
{ |
| 42 |
public: |
| 43 |
typedef RealType value_type; |
| 44 |
|
| 45 |
distribution_archetype(const distribution_archetype&); // Copy constructible. |
| 46 |
distribution_archetype& operator=(const distribution_archetype&); // Assignable. |
| 47 |
|
| 48 |
// There is no default constructor, |
| 49 |
// but we need a way to instantiate the archetype: |
| 50 |
static distribution_archetype& get_object() |
| 51 |
{ |
| 52 |
// will never get called: |
| 53 |
return *reinterpret_cast<distribution_archetype*>(nullptr); |
| 54 |
} |
| 55 |
}; // template <class RealType>class distribution_archetype |
| 56 |
|
| 57 |
// Non-member accessor functions: |
| 58 |
// (This list defines the functions that must be implemented by all distributions). |
| 59 |
|
| 60 |
template <class RealType> |
| 61 |
RealType pdf(const distribution_archetype<RealType>& dist, const RealType& x); |
| 62 |
|
| 63 |
template <class RealType> |
| 64 |
RealType cdf(const distribution_archetype<RealType>& dist, const RealType& x); |
| 65 |
|
| 66 |
template <class RealType> |
| 67 |
RealType quantile(const distribution_archetype<RealType>& dist, const RealType& p); |
| 68 |
|
| 69 |
template <class RealType> |
| 70 |
RealType cdf(const complemented2_type<distribution_archetype<RealType>, RealType>& c); |
| 71 |
|
| 72 |
template <class RealType> |
| 73 |
RealType quantile(const complemented2_type<distribution_archetype<RealType>, RealType>& c); |
| 74 |
|
| 75 |
template <class RealType> |
| 76 |
RealType mean(const distribution_archetype<RealType>& dist); |
| 77 |
|
| 78 |
template <class RealType> |
| 79 |
RealType standard_deviation(const distribution_archetype<RealType>& dist); |
| 80 |
|
| 81 |
template <class RealType> |
| 82 |
RealType variance(const distribution_archetype<RealType>& dist); |
| 83 |
|
| 84 |
template <class RealType> |
| 85 |
RealType hazard(const distribution_archetype<RealType>& dist); |
| 86 |
|
| 87 |
template <class RealType> |
| 88 |
RealType chf(const distribution_archetype<RealType>& dist); |
| 89 |
// http://en.wikipedia.org/wiki/Characteristic_function_%28probability_theory%29 |
| 90 |
|
| 91 |
template <class RealType> |
| 92 |
RealType coefficient_of_variation(const distribution_archetype<RealType>& dist); |
| 93 |
|
| 94 |
template <class RealType> |
| 95 |
RealType mode(const distribution_archetype<RealType>& dist); |
| 96 |
|
| 97 |
template <class RealType> |
| 98 |
RealType skewness(const distribution_archetype<RealType>& dist); |
| 99 |
|
| 100 |
template <class RealType> |
| 101 |
RealType kurtosis_excess(const distribution_archetype<RealType>& dist); |
| 102 |
|
| 103 |
template <class RealType> |
| 104 |
RealType kurtosis(const distribution_archetype<RealType>& dist); |
| 105 |
|
| 106 |
template <class RealType> |
| 107 |
RealType median(const distribution_archetype<RealType>& dist); |
| 108 |
|
| 109 |
template <class RealType> |
| 110 |
std::pair<RealType, RealType> range(const distribution_archetype<RealType>& dist); |
| 111 |
|
| 112 |
template <class RealType> |
| 113 |
std::pair<RealType, RealType> support(const distribution_archetype<RealType>& dist); |
| 114 |
|
| 115 |
// |
| 116 |
// Next comes the concept checks for verifying that a class |
| 117 |
// fulfils the requirements of a Distribution: |
| 118 |
// |
| 119 |
template <class Distribution> |
| 120 |
struct DistributionConcept |
| 121 |
{ |
| 122 |
typedef typename Distribution::value_type value_type; |
| 123 |
|
| 124 |
void constraints() |
| 125 |
{ |
| 126 |
function_requires<CopyConstructibleConcept<Distribution> >(); |
| 127 |
function_requires<AssignableConcept<Distribution> >(); |
| 128 |
|
| 129 |
const Distribution& dist = DistributionConcept<Distribution>::get_object(); |
| 130 |
|
| 131 |
value_type x = 0; |
| 132 |
// The result values are ignored in all these checks. |
| 133 |
value_type v = cdf(dist, x); |
| 134 |
v = cdf(complement(dist, x)); |
| 135 |
suppress_unused_variable_warning(v); |
| 136 |
v = pdf(dist, x); |
| 137 |
suppress_unused_variable_warning(v); |
| 138 |
v = quantile(dist, x); |
| 139 |
suppress_unused_variable_warning(v); |
| 140 |
v = quantile(complement(dist, x)); |
| 141 |
suppress_unused_variable_warning(v); |
| 142 |
v = mean(dist); |
| 143 |
suppress_unused_variable_warning(v); |
| 144 |
v = mode(dist); |
| 145 |
suppress_unused_variable_warning(v); |
| 146 |
v = standard_deviation(dist); |
| 147 |
suppress_unused_variable_warning(v); |
| 148 |
v = variance(dist); |
| 149 |
suppress_unused_variable_warning(v); |
| 150 |
v = hazard(dist, x); |
| 151 |
suppress_unused_variable_warning(v); |
| 152 |
v = chf(dist, x); |
| 153 |
suppress_unused_variable_warning(v); |
| 154 |
v = coefficient_of_variation(dist); |
| 155 |
suppress_unused_variable_warning(v); |
| 156 |
v = skewness(dist); |
| 157 |
suppress_unused_variable_warning(v); |
| 158 |
v = kurtosis(dist); |
| 159 |
suppress_unused_variable_warning(v); |
| 160 |
v = kurtosis_excess(dist); |
| 161 |
suppress_unused_variable_warning(v); |
| 162 |
v = median(dist); |
| 163 |
suppress_unused_variable_warning(v); |
| 164 |
std::pair<value_type, value_type> pv; |
| 165 |
pv = range(dist); |
| 166 |
suppress_unused_variable_warning(pv); |
| 167 |
pv = support(dist); |
| 168 |
suppress_unused_variable_warning(pv); |
| 169 |
|
| 170 |
float f = 1; |
| 171 |
v = cdf(dist, f); |
| 172 |
suppress_unused_variable_warning(v); |
| 173 |
v = cdf(complement(dist, f)); |
| 174 |
suppress_unused_variable_warning(v); |
| 175 |
v = pdf(dist, f); |
| 176 |
suppress_unused_variable_warning(v); |
| 177 |
v = quantile(dist, f); |
| 178 |
suppress_unused_variable_warning(v); |
| 179 |
v = quantile(complement(dist, f)); |
| 180 |
suppress_unused_variable_warning(v); |
| 181 |
v = hazard(dist, f); |
| 182 |
suppress_unused_variable_warning(v); |
| 183 |
v = chf(dist, f); |
| 184 |
suppress_unused_variable_warning(v); |
| 185 |
double d = 1; |
| 186 |
v = cdf(dist, d); |
| 187 |
suppress_unused_variable_warning(v); |
| 188 |
v = cdf(complement(dist, d)); |
| 189 |
suppress_unused_variable_warning(v); |
| 190 |
v = pdf(dist, d); |
| 191 |
suppress_unused_variable_warning(v); |
| 192 |
v = quantile(dist, d); |
| 193 |
suppress_unused_variable_warning(v); |
| 194 |
v = quantile(complement(dist, d)); |
| 195 |
suppress_unused_variable_warning(v); |
| 196 |
v = hazard(dist, d); |
| 197 |
suppress_unused_variable_warning(v); |
| 198 |
v = chf(dist, d); |
| 199 |
suppress_unused_variable_warning(v); |
| 200 |
#ifndef TEST_MPFR |
| 201 |
long double ld = 1; |
| 202 |
v = cdf(dist, ld); |
| 203 |
suppress_unused_variable_warning(v); |
| 204 |
v = cdf(complement(dist, ld)); |
| 205 |
suppress_unused_variable_warning(v); |
| 206 |
v = pdf(dist, ld); |
| 207 |
suppress_unused_variable_warning(v); |
| 208 |
v = quantile(dist, ld); |
| 209 |
suppress_unused_variable_warning(v); |
| 210 |
v = quantile(complement(dist, ld)); |
| 211 |
suppress_unused_variable_warning(v); |
| 212 |
v = hazard(dist, ld); |
| 213 |
suppress_unused_variable_warning(v); |
| 214 |
v = chf(dist, ld); |
| 215 |
suppress_unused_variable_warning(v); |
| 216 |
#endif |
| 217 |
int i = 1; |
| 218 |
v = cdf(dist, i); |
| 219 |
suppress_unused_variable_warning(v); |
| 220 |
v = cdf(complement(dist, i)); |
| 221 |
suppress_unused_variable_warning(v); |
| 222 |
v = pdf(dist, i); |
| 223 |
suppress_unused_variable_warning(v); |
| 224 |
v = quantile(dist, i); |
| 225 |
suppress_unused_variable_warning(v); |
| 226 |
v = quantile(complement(dist, i)); |
| 227 |
suppress_unused_variable_warning(v); |
| 228 |
v = hazard(dist, i); |
| 229 |
suppress_unused_variable_warning(v); |
| 230 |
v = chf(dist, i); |
| 231 |
suppress_unused_variable_warning(v); |
| 232 |
unsigned long li = 1; |
| 233 |
v = cdf(dist, li); |
| 234 |
suppress_unused_variable_warning(v); |
| 235 |
v = cdf(complement(dist, li)); |
| 236 |
suppress_unused_variable_warning(v); |
| 237 |
v = pdf(dist, li); |
| 238 |
suppress_unused_variable_warning(v); |
| 239 |
v = quantile(dist, li); |
| 240 |
suppress_unused_variable_warning(v); |
| 241 |
v = quantile(complement(dist, li)); |
| 242 |
suppress_unused_variable_warning(v); |
| 243 |
v = hazard(dist, li); |
| 244 |
suppress_unused_variable_warning(v); |
| 245 |
v = chf(dist, li); |
| 246 |
suppress_unused_variable_warning(v); |
| 247 |
test_extra_members(dist); |
| 248 |
} |
| 249 |
template <class D> |
| 250 |
static void test_extra_members(const D&) |
| 251 |
{} |
| 252 |
template <class R, class P> |
| 253 |
static void test_extra_members(const boost::math::bernoulli_distribution<R, P>& d) |
| 254 |
{ |
| 255 |
value_type r = d.success_fraction(); |
| 256 |
(void)r; // warning suppression |
| 257 |
} |
| 258 |
template <class R, class P> |
| 259 |
static void test_extra_members(const boost::math::beta_distribution<R, P>& d) |
| 260 |
{ |
| 261 |
value_type r1 = d.alpha(); |
| 262 |
value_type r2 = d.beta(); |
| 263 |
r1 = boost::math::beta_distribution<R, P>::find_alpha(r1, r2); |
| 264 |
suppress_unused_variable_warning(r1); |
| 265 |
r1 = boost::math::beta_distribution<R, P>::find_beta(r1, r2); |
| 266 |
suppress_unused_variable_warning(r1); |
| 267 |
r1 = boost::math::beta_distribution<R, P>::find_alpha(r1, r2, r1); |
| 268 |
suppress_unused_variable_warning(r1); |
| 269 |
r1 = boost::math::beta_distribution<R, P>::find_beta(r1, r2, r1); |
| 270 |
suppress_unused_variable_warning(r1); |
| 271 |
} |
| 272 |
template <class R, class P> |
| 273 |
static void test_extra_members(const boost::math::binomial_distribution<R, P>& d) |
| 274 |
{ |
| 275 |
value_type r = d.success_fraction(); |
| 276 |
r = d.trials(); |
| 277 |
r = Distribution::find_lower_bound_on_p(r, r, r); |
| 278 |
r = Distribution::find_lower_bound_on_p(r, r, r, Distribution::clopper_pearson_exact_interval); |
| 279 |
r = Distribution::find_lower_bound_on_p(r, r, r, Distribution::jeffreys_prior_interval); |
| 280 |
r = Distribution::find_upper_bound_on_p(r, r, r); |
| 281 |
r = Distribution::find_upper_bound_on_p(r, r, r, Distribution::clopper_pearson_exact_interval); |
| 282 |
r = Distribution::find_upper_bound_on_p(r, r, r, Distribution::jeffreys_prior_interval); |
| 283 |
r = Distribution::find_minimum_number_of_trials(r, r, r); |
| 284 |
r = Distribution::find_maximum_number_of_trials(r, r, r); |
| 285 |
suppress_unused_variable_warning(r); |
| 286 |
} |
| 287 |
template <class R, class P> |
| 288 |
static void test_extra_members(const boost::math::cauchy_distribution<R, P>& d) |
| 289 |
{ |
| 290 |
value_type r = d.location(); |
| 291 |
r = d.scale(); |
| 292 |
suppress_unused_variable_warning(r); |
| 293 |
} |
| 294 |
template <class R, class P> |
| 295 |
static void test_extra_members(const boost::math::chi_squared_distribution<R, P>& d) |
| 296 |
{ |
| 297 |
value_type r = d.degrees_of_freedom(); |
| 298 |
r = Distribution::find_degrees_of_freedom(r, r, r, r); |
| 299 |
r = Distribution::find_degrees_of_freedom(r, r, r, r, r); |
| 300 |
suppress_unused_variable_warning(r); |
| 301 |
} |
| 302 |
template <class R, class P> |
| 303 |
static void test_extra_members(const boost::math::exponential_distribution<R, P>& d) |
| 304 |
{ |
| 305 |
value_type r = d.lambda(); |
| 306 |
suppress_unused_variable_warning(r); |
| 307 |
} |
| 308 |
template <class R, class P> |
| 309 |
static void test_extra_members(const boost::math::extreme_value_distribution<R, P>& d) |
| 310 |
{ |
| 311 |
value_type r = d.scale(); |
| 312 |
r = d.location(); |
| 313 |
suppress_unused_variable_warning(r); |
| 314 |
} |
| 315 |
template <class R, class P> |
| 316 |
static void test_extra_members(const boost::math::fisher_f_distribution<R, P>& d) |
| 317 |
{ |
| 318 |
value_type r = d.degrees_of_freedom1(); |
| 319 |
r = d.degrees_of_freedom2(); |
| 320 |
suppress_unused_variable_warning(r); |
| 321 |
} |
| 322 |
template <class R, class P> |
| 323 |
static void test_extra_members(const boost::math::gamma_distribution<R, P>& d) |
| 324 |
{ |
| 325 |
value_type r = d.scale(); |
| 326 |
r = d.shape(); |
| 327 |
suppress_unused_variable_warning(r); |
| 328 |
} |
| 329 |
template <class R, class P> |
| 330 |
static void test_extra_members(const boost::math::inverse_chi_squared_distribution<R, P>& d) |
| 331 |
{ |
| 332 |
value_type r = d.scale(); |
| 333 |
r = d.degrees_of_freedom(); |
| 334 |
suppress_unused_variable_warning(r); |
| 335 |
} |
| 336 |
template <class R, class P> |
| 337 |
static void test_extra_members(const boost::math::inverse_gamma_distribution<R, P>& d) |
| 338 |
{ |
| 339 |
value_type r = d.scale(); |
| 340 |
r = d.shape(); |
| 341 |
suppress_unused_variable_warning(r); |
| 342 |
} |
| 343 |
template <class R, class P> |
| 344 |
static void test_extra_members(const boost::math::hypergeometric_distribution<R, P>& d) |
| 345 |
{ |
| 346 |
std::uint64_t u = d.defective(); |
| 347 |
u = d.sample_count(); |
| 348 |
u = d.total(); |
| 349 |
suppress_unused_variable_warning(u); |
| 350 |
} |
| 351 |
template <class R, class P> |
| 352 |
static void test_extra_members(const boost::math::laplace_distribution<R, P>& d) |
| 353 |
{ |
| 354 |
value_type r = d.scale(); |
| 355 |
r = d.location(); |
| 356 |
suppress_unused_variable_warning(r); |
| 357 |
} |
| 358 |
template <class R, class P> |
| 359 |
static void test_extra_members(const boost::math::logistic_distribution<R, P>& d) |
| 360 |
{ |
| 361 |
value_type r = d.scale(); |
| 362 |
r = d.location(); |
| 363 |
suppress_unused_variable_warning(r); |
| 364 |
} |
| 365 |
template <class R, class P> |
| 366 |
static void test_extra_members(const boost::math::lognormal_distribution<R, P>& d) |
| 367 |
{ |
| 368 |
value_type r = d.scale(); |
| 369 |
r = d.location(); |
| 370 |
suppress_unused_variable_warning(r); |
| 371 |
} |
| 372 |
template <class R, class P> |
| 373 |
static void test_extra_members(const boost::math::negative_binomial_distribution<R, P>& d) |
| 374 |
{ |
| 375 |
value_type r = d.success_fraction(); |
| 376 |
r = d.successes(); |
| 377 |
r = Distribution::find_lower_bound_on_p(r, r, r); |
| 378 |
r = Distribution::find_upper_bound_on_p(r, r, r); |
| 379 |
r = Distribution::find_minimum_number_of_trials(r, r, r); |
| 380 |
r = Distribution::find_maximum_number_of_trials(r, r, r); |
| 381 |
suppress_unused_variable_warning(r); |
| 382 |
} |
| 383 |
template <class R, class P> |
| 384 |
static void test_extra_members(const boost::math::non_central_beta_distribution<R, P>& d) |
| 385 |
{ |
| 386 |
value_type r1 = d.alpha(); |
| 387 |
value_type r2 = d.beta(); |
| 388 |
r1 = d.non_centrality(); |
| 389 |
(void)r1; // warning suppression |
| 390 |
(void)r2; // warning suppression |
| 391 |
} |
| 392 |
template <class R, class P> |
| 393 |
static void test_extra_members(const boost::math::non_central_chi_squared_distribution<R, P>& d) |
| 394 |
{ |
| 395 |
value_type r = d.degrees_of_freedom(); |
| 396 |
r = d.non_centrality(); |
| 397 |
r = Distribution::find_degrees_of_freedom(r, r, r); |
| 398 |
r = Distribution::find_degrees_of_freedom(boost::math::complement(r, r, r)); |
| 399 |
r = Distribution::find_non_centrality(r, r, r); |
| 400 |
r = Distribution::find_non_centrality(boost::math::complement(r, r, r)); |
| 401 |
(void)r; // warning suppression |
| 402 |
} |
| 403 |
template <class R, class P> |
| 404 |
static void test_extra_members(const boost::math::non_central_f_distribution<R, P>& d) |
| 405 |
{ |
| 406 |
value_type r = d.degrees_of_freedom1(); |
| 407 |
r = d.degrees_of_freedom2(); |
| 408 |
r = d.non_centrality(); |
| 409 |
(void)r; // warning suppression |
| 410 |
} |
| 411 |
template <class R, class P> |
| 412 |
static void test_extra_members(const boost::math::non_central_t_distribution<R, P>& d) |
| 413 |
{ |
| 414 |
value_type r = d.degrees_of_freedom(); |
| 415 |
r = d.non_centrality(); |
| 416 |
(void)r; // warning suppression |
| 417 |
} |
| 418 |
template <class R, class P> |
| 419 |
static void test_extra_members(const boost::math::normal_distribution<R, P>& d) |
| 420 |
{ |
| 421 |
value_type r = d.scale(); |
| 422 |
r = d.location(); |
| 423 |
r = d.mean(); |
| 424 |
r = d.standard_deviation(); |
| 425 |
(void)r; // warning suppression |
| 426 |
} |
| 427 |
template <class R, class P> |
| 428 |
static void test_extra_members(const boost::math::pareto_distribution<R, P>& d) |
| 429 |
{ |
| 430 |
value_type r = d.scale(); |
| 431 |
r = d.shape(); |
| 432 |
(void)r; // warning suppression |
| 433 |
} |
| 434 |
template <class R, class P> |
| 435 |
static void test_extra_members(const boost::math::poisson_distribution<R, P>& d) |
| 436 |
{ |
| 437 |
value_type r = d.mean(); |
| 438 |
(void)r; // warning suppression |
| 439 |
} |
| 440 |
template <class R, class P> |
| 441 |
static void test_extra_members(const boost::math::rayleigh_distribution<R, P>& d) |
| 442 |
{ |
| 443 |
value_type r = d.sigma(); |
| 444 |
(void)r; // warning suppression |
| 445 |
} |
| 446 |
template <class R, class P> |
| 447 |
static void test_extra_members(const boost::math::students_t_distribution<R, P>& d) |
| 448 |
{ |
| 449 |
value_type r = d.degrees_of_freedom(); |
| 450 |
r = d.find_degrees_of_freedom(r, r, r, r); |
| 451 |
r = d.find_degrees_of_freedom(r, r, r, r, r); |
| 452 |
(void)r; // warning suppression |
| 453 |
} |
| 454 |
template <class R, class P> |
| 455 |
static void test_extra_members(const boost::math::triangular_distribution<R, P>& d) |
| 456 |
{ |
| 457 |
value_type r = d.lower(); |
| 458 |
r = d.mode(); |
| 459 |
r = d.upper(); |
| 460 |
(void)r; // warning suppression |
| 461 |
} |
| 462 |
template <class R, class P> |
| 463 |
static void test_extra_members(const boost::math::weibull_distribution<R, P>& d) |
| 464 |
{ |
| 465 |
value_type r = d.scale(); |
| 466 |
r = d.shape(); |
| 467 |
(void)r; // warning suppression |
| 468 |
} |
| 469 |
template <class R, class P> |
| 470 |
static void test_extra_members(const boost::math::uniform_distribution<R, P>& d) |
| 471 |
{ |
| 472 |
value_type r = d.lower(); |
| 473 |
r = d.upper(); |
| 474 |
(void)r; // warning suppression |
| 475 |
} |
| 476 |
private: |
| 477 |
static Distribution* pd; |
| 478 |
static Distribution& get_object() |
| 479 |
{ |
| 480 |
// In reality this will never get called: |
| 481 |
return *pd; |
| 482 |
} |
| 483 |
}; // struct DistributionConcept |
| 484 |
|
| 485 |
template <class Distribution> |
| 486 |
Distribution* DistributionConcept<Distribution>::pd = 0; |
| 487 |
|
| 488 |
} // namespace concepts |
| 489 |
} // namespace math |
| 490 |
} // namespace boost |
| 491 |
|
| 492 |
#else |
| 493 |
#error This header can not be used in standalone mode. |
| 494 |
#endif // BOOST_MATH_STANDALONE |
| 495 |
|
| 496 |
#endif // BOOST_MATH_DISTRIBUTION_CONCEPT_HPP |
| 497 |
|