| 1 |
// boost\math\distributions\geometric.hpp |
| 2 |
|
| 3 |
// Copyright John Maddock 2010. |
| 4 |
// Copyright Paul A. Bristow 2010. |
| 5 |
|
| 6 |
// Use, modification and distribution are subject to the |
| 7 |
// Boost Software License, Version 1.0. |
| 8 |
// (See accompanying file LICENSE_1_0.txt |
| 9 |
// or copy at http://www.boost.org/LICENSE_1_0.txt) |
| 10 |
|
| 11 |
// geometric distribution is a discrete probability distribution. |
| 12 |
// It expresses the probability distribution of the number (k) of |
| 13 |
// events, occurrences, failures or arrivals before the first success. |
| 14 |
// supported on the set {0, 1, 2, 3...} |
| 15 |
|
| 16 |
// Note that the set includes zero (unlike some definitions that start at one). |
| 17 |
|
| 18 |
// The random variate k is the number of events, occurrences or arrivals. |
| 19 |
// k argument may be integral, signed, or unsigned, or floating point. |
| 20 |
// If necessary, it has already been promoted from an integral type. |
| 21 |
|
| 22 |
// Note that the geometric distribution |
| 23 |
// (like others including the binomial, geometric & Bernoulli) |
| 24 |
// is strictly defined as a discrete function: |
| 25 |
// only integral values of k are envisaged. |
| 26 |
// However because the method of calculation uses a continuous gamma function, |
| 27 |
// it is convenient to treat it as if a continuous function, |
| 28 |
// and permit non-integral values of k. |
| 29 |
// To enforce the strict mathematical model, users should use floor or ceil functions |
| 30 |
// on k outside this function to ensure that k is integral. |
| 31 |
|
| 32 |
// See http://en.wikipedia.org/wiki/geometric_distribution |
| 33 |
// http://documents.wolfram.com/v5/Add-onsLinks/StandardPackages/Statistics/DiscreteDistributions.html |
| 34 |
// http://mathworld.wolfram.com/GeometricDistribution.html |
| 35 |
|
| 36 |
#ifndef BOOST_MATH_SPECIAL_GEOMETRIC_HPP |
| 37 |
#define BOOST_MATH_SPECIAL_GEOMETRIC_HPP |
| 38 |
|
| 39 |
#include <boost/math/tools/config.hpp> |
| 40 |
#include <boost/math/tools/tuple.hpp> |
| 41 |
#include <boost/math/tools/numeric_limits.hpp> |
| 42 |
#include <boost/math/distributions/fwd.hpp> |
| 43 |
#include <boost/math/special_functions/beta.hpp> // for ibeta(a, b, x) == Ix(a, b). |
| 44 |
#include <boost/math/distributions/complement.hpp> // complement. |
| 45 |
#include <boost/math/distributions/detail/common_error_handling.hpp> // error checks domain_error & logic_error. |
| 46 |
#include <boost/math/special_functions/fpclassify.hpp> // isnan. |
| 47 |
#include <boost/math/tools/roots.hpp> // for root finding. |
| 48 |
#include <boost/math/distributions/detail/inv_discrete_quantile.hpp> |
| 49 |
#include <boost/math/special_functions/log1p.hpp> |
| 50 |
|
| 51 |
#if defined (BOOST_MSVC) |
| 52 |
# pragma warning(push) |
| 53 |
// This believed not now necessary, so commented out. |
| 54 |
//# pragma warning(disable: 4702) // unreachable code. |
| 55 |
// in domain_error_imp in error_handling. |
| 56 |
#endif |
| 57 |
|
| 58 |
namespace boost |
| 59 |
{ |
| 60 |
namespace math |
| 61 |
{ |
| 62 |
namespace geometric_detail |
| 63 |
{ |
| 64 |
// Common error checking routines for geometric distribution function: |
| 65 |
template <class RealType, class Policy> |
| 66 |
BOOST_MATH_GPU_ENABLED inline bool check_success_fraction(const char* function, const RealType& p, RealType* result, const Policy& pol) |
| 67 |
{ |
| 68 |
if( !(boost::math::isfinite)(p) || (p < 0) || (p > 1) ) |
| 69 |
{ |
| 70 |
*result = policies::raise_domain_error<RealType>( |
| 71 |
function, |
| 72 |
"Success fraction argument is %1%, but must be >= 0 and <= 1 !", p, pol); |
| 73 |
return false; |
| 74 |
} |
| 75 |
return true; |
| 76 |
} |
| 77 |
|
| 78 |
template <class RealType, class Policy> |
| 79 |
BOOST_MATH_GPU_ENABLED inline bool check_dist(const char* function, const RealType& p, RealType* result, const Policy& pol) |
| 80 |
{ |
| 81 |
return check_success_fraction(function, p, result, pol); |
| 82 |
} |
| 83 |
|
| 84 |
template <class RealType, class Policy> |
| 85 |
BOOST_MATH_GPU_ENABLED inline bool check_dist_and_k(const char* function, const RealType& p, RealType k, RealType* result, const Policy& pol) |
| 86 |
{ |
| 87 |
if(check_dist(function, p, result, pol) == false) |
| 88 |
{ |
| 89 |
return false; |
| 90 |
} |
| 91 |
if( !(boost::math::isfinite)(k) || (k < 0) ) |
| 92 |
{ // Check k failures. |
| 93 |
*result = policies::raise_domain_error<RealType>( |
| 94 |
function, |
| 95 |
"Number of failures argument is %1%, but must be >= 0 !", k, pol); |
| 96 |
return false; |
| 97 |
} |
| 98 |
return true; |
| 99 |
} // Check_dist_and_k |
| 100 |
|
| 101 |
template <class RealType, class Policy> |
| 102 |
BOOST_MATH_GPU_ENABLED inline bool check_dist_and_prob(const char* function, RealType p, RealType prob, RealType* result, const Policy& pol) |
| 103 |
{ |
| 104 |
if((check_dist(function, p, result, pol) && detail::check_probability(function, prob, result, pol)) == false) |
| 105 |
{ |
| 106 |
return false; |
| 107 |
} |
| 108 |
return true; |
| 109 |
} // check_dist_and_prob |
| 110 |
} // namespace geometric_detail |
| 111 |
|
| 112 |
template <class RealType = double, class Policy = policies::policy<> > |
| 113 |
class geometric_distribution |
| 114 |
{ |
| 115 |
public: |
| 116 |
typedef RealType value_type; |
| 117 |
typedef Policy policy_type; |
| 118 |
|
| 119 |
BOOST_MATH_GPU_ENABLED geometric_distribution(RealType p) : m_p(p) |
| 120 |
{ // Constructor stores success_fraction p. |
| 121 |
RealType result; |
| 122 |
geometric_detail::check_dist( |
| 123 |
"geometric_distribution<%1%>::geometric_distribution", |
| 124 |
m_p, // Check success_fraction 0 <= p <= 1. |
| 125 |
&result, Policy()); |
| 126 |
} // geometric_distribution constructor. |
| 127 |
|
| 128 |
// Private data getter class member functions. |
| 129 |
BOOST_MATH_GPU_ENABLED RealType success_fraction() const |
| 130 |
{ // Probability of success as fraction in range 0 to 1. |
| 131 |
return m_p; |
| 132 |
} |
| 133 |
BOOST_MATH_GPU_ENABLED RealType successes() const |
| 134 |
{ // Total number of successes r = 1 (for compatibility with negative binomial?). |
| 135 |
return 1; |
| 136 |
} |
| 137 |
|
| 138 |
// Parameter estimation. |
| 139 |
// (These are copies of negative_binomial distribution with successes = 1). |
| 140 |
BOOST_MATH_GPU_ENABLED static RealType find_lower_bound_on_p( |
| 141 |
RealType trials, |
| 142 |
RealType alpha) // alpha 0.05 equivalent to 95% for one-sided test. |
| 143 |
{ |
| 144 |
constexpr auto function = "boost::math::geometric<%1%>::find_lower_bound_on_p"; |
| 145 |
RealType result = 0; // of error checks. |
| 146 |
RealType successes = 1; |
| 147 |
RealType failures = trials - successes; |
| 148 |
if(false == detail::check_probability(function, alpha, &result, Policy()) |
| 149 |
&& geometric_detail::check_dist_and_k( |
| 150 |
function, RealType(0), failures, &result, Policy())) |
| 151 |
{ |
| 152 |
return result; |
| 153 |
} |
| 154 |
// Use complement ibeta_inv function for lower bound. |
| 155 |
// This is adapted from the corresponding binomial formula |
| 156 |
// here: http://www.itl.nist.gov/div898/handbook/prc/section2/prc241.htm |
| 157 |
// This is a Clopper-Pearson interval, and may be overly conservative, |
| 158 |
// see also "A Simple Improved Inferential Method for Some |
| 159 |
// Discrete Distributions" Yong CAI and K. KRISHNAMOORTHY |
| 160 |
// http://www.ucs.louisiana.edu/~kxk4695/Discrete_new.pdf |
| 161 |
// |
| 162 |
return ibeta_inv(successes, failures + 1, alpha, static_cast<RealType*>(nullptr), Policy()); |
| 163 |
} // find_lower_bound_on_p |
| 164 |
|
| 165 |
BOOST_MATH_GPU_ENABLED static RealType find_upper_bound_on_p( |
| 166 |
RealType trials, |
| 167 |
RealType alpha) // alpha 0.05 equivalent to 95% for one-sided test. |
| 168 |
{ |
| 169 |
constexpr auto function = "boost::math::geometric<%1%>::find_upper_bound_on_p"; |
| 170 |
RealType result = 0; // of error checks. |
| 171 |
RealType successes = 1; |
| 172 |
RealType failures = trials - successes; |
| 173 |
if(false == geometric_detail::check_dist_and_k( |
| 174 |
function, RealType(0), failures, &result, Policy()) |
| 175 |
&& detail::check_probability(function, alpha, &result, Policy())) |
| 176 |
{ |
| 177 |
return result; |
| 178 |
} |
| 179 |
if(failures == 0) |
| 180 |
{ |
| 181 |
return 1; |
| 182 |
}// Use complement ibetac_inv function for upper bound. |
| 183 |
// Note adjusted failures value: *not* failures+1 as usual. |
| 184 |
// This is adapted from the corresponding binomial formula |
| 185 |
// here: http://www.itl.nist.gov/div898/handbook/prc/section2/prc241.htm |
| 186 |
// This is a Clopper-Pearson interval, and may be overly conservative, |
| 187 |
// see also "A Simple Improved Inferential Method for Some |
| 188 |
// Discrete Distributions" Yong CAI and K. Krishnamoorthy |
| 189 |
// http://www.ucs.louisiana.edu/~kxk4695/Discrete_new.pdf |
| 190 |
// |
| 191 |
return ibetac_inv(successes, failures, alpha, static_cast<RealType*>(nullptr), Policy()); |
| 192 |
} // find_upper_bound_on_p |
| 193 |
|
| 194 |
// Estimate number of trials : |
| 195 |
// "How many trials do I need to be P% sure of seeing k or fewer failures?" |
| 196 |
|
| 197 |
BOOST_MATH_GPU_ENABLED static RealType find_minimum_number_of_trials( |
| 198 |
RealType k, // number of failures (k >= 0). |
| 199 |
RealType p, // success fraction 0 <= p <= 1. |
| 200 |
RealType alpha) // risk level threshold 0 <= alpha <= 1. |
| 201 |
{ |
| 202 |
constexpr auto function = "boost::math::geometric<%1%>::find_minimum_number_of_trials"; |
| 203 |
// Error checks: |
| 204 |
RealType result = 0; |
| 205 |
if(false == geometric_detail::check_dist_and_k( |
| 206 |
function, p, k, &result, Policy()) |
| 207 |
&& detail::check_probability(function, alpha, &result, Policy())) |
| 208 |
{ |
| 209 |
return result; |
| 210 |
} |
| 211 |
result = ibeta_inva(k + 1, p, alpha, Policy()); // returns n - k |
| 212 |
return result + k; |
| 213 |
} // RealType find_number_of_failures |
| 214 |
|
| 215 |
BOOST_MATH_GPU_ENABLED static RealType find_maximum_number_of_trials( |
| 216 |
RealType k, // number of failures (k >= 0). |
| 217 |
RealType p, // success fraction 0 <= p <= 1. |
| 218 |
RealType alpha) // risk level threshold 0 <= alpha <= 1. |
| 219 |
{ |
| 220 |
constexpr auto function = "boost::math::geometric<%1%>::find_maximum_number_of_trials"; |
| 221 |
// Error checks: |
| 222 |
RealType result = 0; |
| 223 |
if(false == geometric_detail::check_dist_and_k( |
| 224 |
function, p, k, &result, Policy()) |
| 225 |
&& detail::check_probability(function, alpha, &result, Policy())) |
| 226 |
{ |
| 227 |
return result; |
| 228 |
} |
| 229 |
result = ibetac_inva(k + 1, p, alpha, Policy()); // returns n - k |
| 230 |
return result + k; |
| 231 |
} // RealType find_number_of_trials complemented |
| 232 |
|
| 233 |
private: |
| 234 |
//RealType m_r; // successes fixed at unity. |
| 235 |
RealType m_p; // success_fraction |
| 236 |
}; // template <class RealType, class Policy> class geometric_distribution |
| 237 |
|
| 238 |
typedef geometric_distribution<double> geometric; // Reserved name of type double. |
| 239 |
|
| 240 |
#ifdef __cpp_deduction_guides |
| 241 |
template <class RealType> |
| 242 |
geometric_distribution(RealType)->geometric_distribution<typename boost::math::tools::promote_args<RealType>::type>; |
| 243 |
#endif |
| 244 |
|
| 245 |
template <class RealType, class Policy> |
| 246 |
BOOST_MATH_GPU_ENABLED inline const boost::math::pair<RealType, RealType> range(const geometric_distribution<RealType, Policy>& /* dist */) |
| 247 |
{ // Range of permissible values for random variable k. |
| 248 |
using boost::math::tools::max_value; |
| 249 |
return boost::math::pair<RealType, RealType>(static_cast<RealType>(0), max_value<RealType>()); // max_integer? |
| 250 |
} |
| 251 |
|
| 252 |
template <class RealType, class Policy> |
| 253 |
BOOST_MATH_GPU_ENABLED inline const boost::math::pair<RealType, RealType> support(const geometric_distribution<RealType, Policy>& /* dist */) |
| 254 |
{ // Range of supported values for random variable k. |
| 255 |
// This is range where cdf rises from 0 to 1, and outside it, the pdf is zero. |
| 256 |
using boost::math::tools::max_value; |
| 257 |
return boost::math::pair<RealType, RealType>(static_cast<RealType>(0), max_value<RealType>()); // max_integer? |
| 258 |
} |
| 259 |
|
| 260 |
template <class RealType, class Policy> |
| 261 |
BOOST_MATH_GPU_ENABLED inline RealType mean(const geometric_distribution<RealType, Policy>& dist) |
| 262 |
{ // Mean of geometric distribution = (1-p)/p. |
| 263 |
return (1 - dist.success_fraction() ) / dist.success_fraction(); |
| 264 |
} // mean |
| 265 |
|
| 266 |
// median implemented via quantile(half) in derived accessors. |
| 267 |
|
| 268 |
template <class RealType, class Policy> |
| 269 |
BOOST_MATH_GPU_ENABLED inline RealType mode(const geometric_distribution<RealType, Policy>&) |
| 270 |
{ // Mode of geometric distribution = zero. |
| 271 |
BOOST_MATH_STD_USING // ADL of std functions. |
| 272 |
return 0; |
| 273 |
} // mode |
| 274 |
|
| 275 |
template <class RealType, class Policy> |
| 276 |
BOOST_MATH_GPU_ENABLED inline RealType variance(const geometric_distribution<RealType, Policy>& dist) |
| 277 |
{ // Variance of Binomial distribution = (1-p) / p^2. |
| 278 |
return (1 - dist.success_fraction()) |
| 279 |
/ (dist.success_fraction() * dist.success_fraction()); |
| 280 |
} // variance |
| 281 |
|
| 282 |
template <class RealType, class Policy> |
| 283 |
BOOST_MATH_GPU_ENABLED inline RealType skewness(const geometric_distribution<RealType, Policy>& dist) |
| 284 |
{ // skewness of geometric distribution = 2-p / (sqrt(r(1-p)) |
| 285 |
BOOST_MATH_STD_USING // ADL of std functions. |
| 286 |
RealType p = dist.success_fraction(); |
| 287 |
return (2 - p) / sqrt(1 - p); |
| 288 |
} // skewness |
| 289 |
|
| 290 |
template <class RealType, class Policy> |
| 291 |
BOOST_MATH_GPU_ENABLED inline RealType kurtosis(const geometric_distribution<RealType, Policy>& dist) |
| 292 |
{ // kurtosis of geometric distribution |
| 293 |
// http://en.wikipedia.org/wiki/geometric is kurtosis_excess so add 3 |
| 294 |
RealType p = dist.success_fraction(); |
| 295 |
return 3 + (p*p - 6*p + 6) / (1 - p); |
| 296 |
} // kurtosis |
| 297 |
|
| 298 |
template <class RealType, class Policy> |
| 299 |
BOOST_MATH_GPU_ENABLED inline RealType kurtosis_excess(const geometric_distribution<RealType, Policy>& dist) |
| 300 |
{ // kurtosis excess of geometric distribution |
| 301 |
// http://mathworld.wolfram.com/Kurtosis.html table of kurtosis_excess |
| 302 |
RealType p = dist.success_fraction(); |
| 303 |
return (p*p - 6*p + 6) / (1 - p); |
| 304 |
} // kurtosis_excess |
| 305 |
|
| 306 |
// RealType standard_deviation(const geometric_distribution<RealType, Policy>& dist) |
| 307 |
// standard_deviation provided by derived accessors. |
| 308 |
// RealType hazard(const geometric_distribution<RealType, Policy>& dist) |
| 309 |
// hazard of geometric distribution provided by derived accessors. |
| 310 |
// RealType chf(const geometric_distribution<RealType, Policy>& dist) |
| 311 |
// chf of geometric distribution provided by derived accessors. |
| 312 |
|
| 313 |
template <class RealType, class Policy> |
| 314 |
BOOST_MATH_GPU_ENABLED inline RealType pdf(const geometric_distribution<RealType, Policy>& dist, const RealType& k) |
| 315 |
{ // Probability Density/Mass Function. |
| 316 |
BOOST_FPU_EXCEPTION_GUARD |
| 317 |
BOOST_MATH_STD_USING // For ADL of math functions. |
| 318 |
constexpr auto function = "boost::math::pdf(const geometric_distribution<%1%>&, %1%)"; |
| 319 |
|
| 320 |
RealType p = dist.success_fraction(); |
| 321 |
RealType result = 0; |
| 322 |
if(false == geometric_detail::check_dist_and_k( |
| 323 |
function, |
| 324 |
p, |
| 325 |
k, |
| 326 |
&result, Policy())) |
| 327 |
{ |
| 328 |
return result; |
| 329 |
} |
| 330 |
if (k == 0) |
| 331 |
{ |
| 332 |
return p; // success_fraction |
| 333 |
} |
| 334 |
RealType q = 1 - p; // Inaccurate for small p? |
| 335 |
// So try to avoid inaccuracy for large or small p. |
| 336 |
// but has little effect > last significant bit. |
| 337 |
//cout << "p * pow(q, k) " << result << endl; // seems best whatever p |
| 338 |
//cout << "exp(p * k * log1p(-p)) " << p * exp(k * log1p(-p)) << endl; |
| 339 |
//if (p < 0.5) |
| 340 |
//{ |
| 341 |
// result = p * pow(q, k); |
| 342 |
//} |
| 343 |
//else |
| 344 |
//{ |
| 345 |
// result = p * exp(k * log1p(-p)); |
| 346 |
//} |
| 347 |
result = p * pow(q, k); |
| 348 |
return result; |
| 349 |
} // geometric_pdf |
| 350 |
|
| 351 |
template <class RealType, class Policy> |
| 352 |
BOOST_MATH_GPU_ENABLED inline RealType cdf(const geometric_distribution<RealType, Policy>& dist, const RealType& k) |
| 353 |
{ // Cumulative Distribution Function of geometric. |
| 354 |
constexpr auto function = "boost::math::cdf(const geometric_distribution<%1%>&, %1%)"; |
| 355 |
|
| 356 |
// k argument may be integral, signed, or unsigned, or floating point. |
| 357 |
// If necessary, it has already been promoted from an integral type. |
| 358 |
RealType p = dist.success_fraction(); |
| 359 |
// Error check: |
| 360 |
RealType result = 0; |
| 361 |
if(false == geometric_detail::check_dist_and_k( |
| 362 |
function, |
| 363 |
p, |
| 364 |
k, |
| 365 |
&result, Policy())) |
| 366 |
{ |
| 367 |
return result; |
| 368 |
} |
| 369 |
if(k == 0) |
| 370 |
{ |
| 371 |
return p; // success_fraction |
| 372 |
} |
| 373 |
//RealType q = 1 - p; // Bad for small p |
| 374 |
//RealType probability = 1 - std::pow(q, k+1); |
| 375 |
|
| 376 |
RealType z = boost::math::log1p(-p, Policy()) * (k + 1); |
| 377 |
RealType probability = -boost::math::expm1(z, Policy()); |
| 378 |
|
| 379 |
return probability; |
| 380 |
} // cdf Cumulative Distribution Function geometric. |
| 381 |
|
| 382 |
template <class RealType, class Policy> |
| 383 |
BOOST_MATH_GPU_ENABLED inline RealType logcdf(const geometric_distribution<RealType, Policy>& dist, const RealType& k) |
| 384 |
{ // Cumulative Distribution Function of geometric. |
| 385 |
BOOST_MATH_STD_USING |
| 386 |
constexpr auto function = "boost::math::logcdf(const geometric_distribution<%1%>&, %1%)"; |
| 387 |
|
| 388 |
// k argument may be integral, signed, or unsigned, or floating point. |
| 389 |
// If necessary, it has already been promoted from an integral type. |
| 390 |
RealType p = dist.success_fraction(); |
| 391 |
// Error check: |
| 392 |
RealType result = 0; |
| 393 |
if(false == geometric_detail::check_dist_and_k( |
| 394 |
function, |
| 395 |
p, |
| 396 |
k, |
| 397 |
&result, Policy())) |
| 398 |
{ |
| 399 |
return -boost::math::numeric_limits<RealType>::infinity(); |
| 400 |
} |
| 401 |
if(k == 0) |
| 402 |
{ |
| 403 |
return log(p); // success_fraction |
| 404 |
} |
| 405 |
//RealType q = 1 - p; // Bad for small p |
| 406 |
//RealType probability = 1 - std::pow(q, k+1); |
| 407 |
|
| 408 |
RealType z = boost::math::log1p(-p, Policy()) * (k + 1); |
| 409 |
return log1p(-exp(z), Policy()); |
| 410 |
} // logcdf Cumulative Distribution Function geometric. |
| 411 |
|
| 412 |
template <class RealType, class Policy> |
| 413 |
BOOST_MATH_GPU_ENABLED inline RealType cdf(const complemented2_type<geometric_distribution<RealType, Policy>, RealType>& c) |
| 414 |
{ // Complemented Cumulative Distribution Function geometric. |
| 415 |
BOOST_MATH_STD_USING |
| 416 |
constexpr auto function = "boost::math::cdf(const geometric_distribution<%1%>&, %1%)"; |
| 417 |
// k argument may be integral, signed, or unsigned, or floating point. |
| 418 |
// If necessary, it has already been promoted from an integral type. |
| 419 |
RealType const& k = c.param; |
| 420 |
geometric_distribution<RealType, Policy> const& dist = c.dist; |
| 421 |
RealType p = dist.success_fraction(); |
| 422 |
// Error check: |
| 423 |
RealType result = 0; |
| 424 |
if(false == geometric_detail::check_dist_and_k( |
| 425 |
function, |
| 426 |
p, |
| 427 |
k, |
| 428 |
&result, Policy())) |
| 429 |
{ |
| 430 |
return result; |
| 431 |
} |
| 432 |
RealType z = boost::math::log1p(-p, Policy()) * (k+1); |
| 433 |
RealType probability = exp(z); |
| 434 |
return probability; |
| 435 |
} // cdf Complemented Cumulative Distribution Function geometric. |
| 436 |
|
| 437 |
template <class RealType, class Policy> |
| 438 |
BOOST_MATH_GPU_ENABLED inline RealType logcdf(const complemented2_type<geometric_distribution<RealType, Policy>, RealType>& c) |
| 439 |
{ // Complemented Cumulative Distribution Function geometric. |
| 440 |
BOOST_MATH_STD_USING |
| 441 |
constexpr auto function = "boost::math::logcdf(const geometric_distribution<%1%>&, %1%)"; |
| 442 |
// k argument may be integral, signed, or unsigned, or floating point. |
| 443 |
// If necessary, it has already been promoted from an integral type. |
| 444 |
RealType const& k = c.param; |
| 445 |
geometric_distribution<RealType, Policy> const& dist = c.dist; |
| 446 |
RealType p = dist.success_fraction(); |
| 447 |
// Error check: |
| 448 |
RealType result = 0; |
| 449 |
if(false == geometric_detail::check_dist_and_k( |
| 450 |
function, |
| 451 |
p, |
| 452 |
k, |
| 453 |
&result, Policy())) |
| 454 |
{ |
| 455 |
return -boost::math::numeric_limits<RealType>::infinity(); |
| 456 |
} |
| 457 |
|
| 458 |
return boost::math::log1p(-p, Policy()) * (k+1); |
| 459 |
} // logcdf Complemented Cumulative Distribution Function geometric. |
| 460 |
|
| 461 |
template <class RealType, class Policy> |
| 462 |
BOOST_MATH_GPU_ENABLED inline RealType quantile(const geometric_distribution<RealType, Policy>& dist, const RealType& x) |
| 463 |
{ // Quantile, percentile/100 or Percent Point geometric function. |
| 464 |
// Return the number of expected failures k for a given probability p. |
| 465 |
|
| 466 |
// Inverse cumulative Distribution Function or Quantile (percentile / 100) of geometric Probability. |
| 467 |
// k argument may be integral, signed, or unsigned, or floating point. |
| 468 |
|
| 469 |
constexpr auto function = "boost::math::quantile(const geometric_distribution<%1%>&, %1%)"; |
| 470 |
BOOST_MATH_STD_USING // ADL of std functions. |
| 471 |
|
| 472 |
RealType success_fraction = dist.success_fraction(); |
| 473 |
// Check dist and x. |
| 474 |
RealType result = 0; |
| 475 |
if(false == geometric_detail::check_dist_and_prob |
| 476 |
(function, success_fraction, x, &result, Policy())) |
| 477 |
{ |
| 478 |
return result; |
| 479 |
} |
| 480 |
|
| 481 |
// Special cases. |
| 482 |
if (x == 1) |
| 483 |
{ // Would need +infinity failures for total confidence. |
| 484 |
result = policies::raise_overflow_error<RealType>( |
| 485 |
function, |
| 486 |
"Probability argument is 1, which implies infinite failures !", Policy()); |
| 487 |
return result; |
| 488 |
// usually means return +std::numeric_limits<RealType>::infinity(); |
| 489 |
// unless #define BOOST_MATH_THROW_ON_OVERFLOW_ERROR |
| 490 |
} |
| 491 |
if (x == 0) |
| 492 |
{ // No failures are expected if P = 0. |
| 493 |
return 0; // Total trials will be just dist.successes. |
| 494 |
} |
| 495 |
// if (P <= pow(dist.success_fraction(), 1)) |
| 496 |
if (x <= success_fraction) |
| 497 |
{ // p <= pdf(dist, 0) == cdf(dist, 0) |
| 498 |
return 0; |
| 499 |
} |
| 500 |
if (x == 1) |
| 501 |
{ |
| 502 |
return 0; |
| 503 |
} |
| 504 |
|
| 505 |
// log(1-x) /log(1-success_fraction) -1; but use log1p in case success_fraction is small |
| 506 |
result = boost::math::log1p(-x, Policy()) / boost::math::log1p(-success_fraction, Policy()) - 1; |
| 507 |
// Subtract a few epsilons here too? |
| 508 |
// to make sure it doesn't slip over, so ceil would be one too many. |
| 509 |
return result; |
| 510 |
} // RealType quantile(const geometric_distribution dist, p) |
| 511 |
|
| 512 |
template <class RealType, class Policy> |
| 513 |
BOOST_MATH_GPU_ENABLED inline RealType quantile(const complemented2_type<geometric_distribution<RealType, Policy>, RealType>& c) |
| 514 |
{ // Quantile or Percent Point Binomial function. |
| 515 |
// Return the number of expected failures k for a given |
| 516 |
// complement of the probability Q = 1 - P. |
| 517 |
constexpr auto function = "boost::math::quantile(const geometric_distribution<%1%>&, %1%)"; |
| 518 |
BOOST_MATH_STD_USING |
| 519 |
// Error checks: |
| 520 |
RealType x = c.param; |
| 521 |
const geometric_distribution<RealType, Policy>& dist = c.dist; |
| 522 |
RealType success_fraction = dist.success_fraction(); |
| 523 |
RealType result = 0; |
| 524 |
if(false == geometric_detail::check_dist_and_prob( |
| 525 |
function, |
| 526 |
success_fraction, |
| 527 |
x, |
| 528 |
&result, Policy())) |
| 529 |
{ |
| 530 |
return result; |
| 531 |
} |
| 532 |
|
| 533 |
// Special cases: |
| 534 |
if(x == 1) |
| 535 |
{ // There may actually be no answer to this question, |
| 536 |
// since the probability of zero failures may be non-zero, |
| 537 |
return 0; // but zero is the best we can do: |
| 538 |
} |
| 539 |
if (-x <= boost::math::powm1(dist.success_fraction(), dist.successes(), Policy())) |
| 540 |
{ // q <= cdf(complement(dist, 0)) == pdf(dist, 0) |
| 541 |
return 0; // |
| 542 |
} |
| 543 |
if(x == 0) |
| 544 |
{ // Probability 1 - Q == 1 so infinite failures to achieve certainty. |
| 545 |
// Would need +infinity failures for total confidence. |
| 546 |
result = policies::raise_overflow_error<RealType>( |
| 547 |
function, |
| 548 |
"Probability argument complement is 0, which implies infinite failures !", Policy()); |
| 549 |
return result; |
| 550 |
// usually means return +std::numeric_limits<RealType>::infinity(); |
| 551 |
// unless #define BOOST_MATH_THROW_ON_OVERFLOW_ERROR |
| 552 |
} |
| 553 |
// log(x) /log(1-success_fraction) -1; but use log1p in case success_fraction is small |
| 554 |
result = log(x) / boost::math::log1p(-success_fraction, Policy()) - 1; |
| 555 |
return result; |
| 556 |
|
| 557 |
} // quantile complement |
| 558 |
|
| 559 |
} // namespace math |
| 560 |
} // namespace boost |
| 561 |
|
| 562 |
// This include must be at the end, *after* the accessors |
| 563 |
// for this distribution have been defined, in order to |
| 564 |
// keep compilers that support two-phase lookup happy. |
| 565 |
#include <boost/math/distributions/detail/derived_accessors.hpp> |
| 566 |
|
| 567 |
#if defined (BOOST_MSVC) |
| 568 |
# pragma warning(pop) |
| 569 |
#endif |
| 570 |
|
| 571 |
#endif // BOOST_MATH_SPECIAL_GEOMETRIC_HPP |