| 1 | // Copyright John Maddock 2006. |
| 2 | |
| 3 | // Use, modification and distribution are subject to the |
| 4 | // Boost Software License, Version 1.0. |
| 5 | // (See accompanying file LICENSE_1_0.txt |
| 6 | // or copy at http://www.boost.org/LICENSE_1_0.txt) |
| 7 | |
| 8 | #ifndef BOOST_MATH_DISTRIBUTIONS_FISHER_F_HPP |
| 9 | #define BOOST_MATH_DISTRIBUTIONS_FISHER_F_HPP |
| 10 | |
| 11 | #include <boost/math/distributions/fwd.hpp> |
| 12 | #include <boost/math/special_functions/beta.hpp> // for incomplete beta. |
| 13 | #include <boost/math/distributions/complement.hpp> // complements |
| 14 | #include <boost/math/distributions/detail/common_error_handling.hpp> // error checks |
| 15 | #include <boost/math/special_functions/fpclassify.hpp> |
| 16 | |
| 17 | #include <utility> |
| 18 | |
| 19 | namespace boost{ namespace math{ |
| 20 | |
| 21 | template <class RealType = double, class Policy = policies::policy<> > |
| 22 | class fisher_f_distribution |
| 23 | { |
| 24 | public: |
| 25 | typedef RealType value_type; |
| 26 | typedef Policy policy_type; |
| 27 | |
| 28 | fisher_f_distribution(const RealType& i, const RealType& j) : m_df1(i), m_df2(j) |
| 29 | { |
| 30 | static const char* function = "fisher_f_distribution<%1%>::fisher_f_distribution" ; |
| 31 | RealType result; |
| 32 | detail::check_df( |
| 33 | function, m_df1, &result, Policy()); |
| 34 | detail::check_df( |
| 35 | function, m_df2, &result, Policy()); |
| 36 | } // fisher_f_distribution |
| 37 | |
| 38 | RealType degrees_of_freedom1()const |
| 39 | { |
| 40 | return m_df1; |
| 41 | } |
| 42 | RealType degrees_of_freedom2()const |
| 43 | { |
| 44 | return m_df2; |
| 45 | } |
| 46 | |
| 47 | private: |
| 48 | // |
| 49 | // Data members: |
| 50 | // |
| 51 | RealType m_df1; // degrees of freedom are a real number. |
| 52 | RealType m_df2; // degrees of freedom are a real number. |
| 53 | }; |
| 54 | |
| 55 | typedef fisher_f_distribution<double> fisher_f; |
| 56 | |
| 57 | #ifdef __cpp_deduction_guides |
| 58 | template <class RealType> |
| 59 | fisher_f_distribution(RealType,RealType)->fisher_f_distribution<typename boost::math::tools::promote_args<RealType>::type>; |
| 60 | #endif |
| 61 | |
| 62 | template <class RealType, class Policy> |
| 63 | inline const std::pair<RealType, RealType> range(const fisher_f_distribution<RealType, Policy>& /*dist*/) |
| 64 | { // Range of permissible values for random variable x. |
| 65 | using boost::math::tools::max_value; |
| 66 | return std::pair<RealType, RealType>(static_cast<RealType>(0), max_value<RealType>()); |
| 67 | } |
| 68 | |
| 69 | template <class RealType, class Policy> |
| 70 | inline const std::pair<RealType, RealType> support(const fisher_f_distribution<RealType, Policy>& /*dist*/) |
| 71 | { // Range of supported values for random variable x. |
| 72 | // This is range where cdf rises from 0 to 1, and outside it, the pdf is zero. |
| 73 | using boost::math::tools::max_value; |
| 74 | return std::pair<RealType, RealType>(static_cast<RealType>(0), max_value<RealType>()); |
| 75 | } |
| 76 | |
| 77 | template <class RealType, class Policy> |
| 78 | RealType pdf(const fisher_f_distribution<RealType, Policy>& dist, const RealType& x) |
| 79 | { |
| 80 | BOOST_MATH_STD_USING // for ADL of std functions |
| 81 | RealType df1 = dist.degrees_of_freedom1(); |
| 82 | RealType df2 = dist.degrees_of_freedom2(); |
| 83 | // Error check: |
| 84 | RealType error_result = 0; |
| 85 | static const char* function = "boost::math::pdf(fisher_f_distribution<%1%> const&, %1%)" ; |
| 86 | if(false == (detail::check_df( |
| 87 | function, df1, &error_result, Policy()) |
| 88 | && detail::check_df( |
| 89 | function, df2, &error_result, Policy()))) |
| 90 | return error_result; |
| 91 | |
| 92 | if((x < 0) || !(boost::math::isfinite)(x)) |
| 93 | { |
| 94 | return policies::raise_domain_error<RealType>( |
| 95 | function, "Random variable parameter was %1%, but must be > 0 !" , x, Policy()); |
| 96 | } |
| 97 | |
| 98 | if(x == 0) |
| 99 | { |
| 100 | // special cases: |
| 101 | if(df1 < 2) |
| 102 | return policies::raise_overflow_error<RealType>( |
| 103 | function, 0, Policy()); |
| 104 | else if(df1 == 2) |
| 105 | return 1; |
| 106 | else |
| 107 | return 0; |
| 108 | } |
| 109 | |
| 110 | // |
| 111 | // You reach this formula by direct differentiation of the |
| 112 | // cdf expressed in terms of the incomplete beta. |
| 113 | // |
| 114 | // There are two versions so we don't pass a value of z |
| 115 | // that is very close to 1 to ibeta_derivative: for some values |
| 116 | // of df1 and df2, all the change takes place in this area. |
| 117 | // |
| 118 | RealType v1x = df1 * x; |
| 119 | RealType result; |
| 120 | if(v1x > df2) |
| 121 | { |
| 122 | result = (df2 * df1) / ((df2 + v1x) * (df2 + v1x)); |
| 123 | result *= ibeta_derivative(df2 / 2, df1 / 2, df2 / (df2 + v1x), Policy()); |
| 124 | } |
| 125 | else |
| 126 | { |
| 127 | result = df2 + df1 * x; |
| 128 | result = (result * df1 - x * df1 * df1) / (result * result); |
| 129 | result *= ibeta_derivative(df1 / 2, df2 / 2, v1x / (df2 + v1x), Policy()); |
| 130 | } |
| 131 | return result; |
| 132 | } // pdf |
| 133 | |
| 134 | template <class RealType, class Policy> |
| 135 | inline RealType cdf(const fisher_f_distribution<RealType, Policy>& dist, const RealType& x) |
| 136 | { |
| 137 | static const char* function = "boost::math::cdf(fisher_f_distribution<%1%> const&, %1%)" ; |
| 138 | RealType df1 = dist.degrees_of_freedom1(); |
| 139 | RealType df2 = dist.degrees_of_freedom2(); |
| 140 | // Error check: |
| 141 | RealType error_result = 0; |
| 142 | if(false == detail::check_df( |
| 143 | function, df1, &error_result, Policy()) |
| 144 | && detail::check_df( |
| 145 | function, df2, &error_result, Policy())) |
| 146 | return error_result; |
| 147 | |
| 148 | if((x < 0) || !(boost::math::isfinite)(x)) |
| 149 | { |
| 150 | return policies::raise_domain_error<RealType>( |
| 151 | function, "Random Variable parameter was %1%, but must be > 0 !" , x, Policy()); |
| 152 | } |
| 153 | |
| 154 | RealType v1x = df1 * x; |
| 155 | // |
| 156 | // There are two equivalent formulas used here, the aim is |
| 157 | // to prevent the final argument to the incomplete beta |
| 158 | // from being too close to 1: for some values of df1 and df2 |
| 159 | // the rate of change can be arbitrarily large in this area, |
| 160 | // whilst the value we're passing will have lost information |
| 161 | // content as a result of being 0.999999something. Better |
| 162 | // to switch things around so we're passing 1-z instead. |
| 163 | // |
| 164 | return v1x > df2 |
| 165 | ? boost::math::ibetac(df2 / 2, df1 / 2, df2 / (df2 + v1x), Policy()) |
| 166 | : boost::math::ibeta(df1 / 2, df2 / 2, v1x / (df2 + v1x), Policy()); |
| 167 | } // cdf |
| 168 | |
| 169 | template <class RealType, class Policy> |
| 170 | inline RealType quantile(const fisher_f_distribution<RealType, Policy>& dist, const RealType& p) |
| 171 | { |
| 172 | static const char* function = "boost::math::quantile(fisher_f_distribution<%1%> const&, %1%)" ; |
| 173 | RealType df1 = dist.degrees_of_freedom1(); |
| 174 | RealType df2 = dist.degrees_of_freedom2(); |
| 175 | // Error check: |
| 176 | RealType error_result = 0; |
| 177 | if(false == (detail::check_df( |
| 178 | function, df1, &error_result, Policy()) |
| 179 | && detail::check_df( |
| 180 | function, df2, &error_result, Policy()) |
| 181 | && detail::check_probability( |
| 182 | function, p, &error_result, Policy()))) |
| 183 | return error_result; |
| 184 | |
| 185 | // With optimizations turned on, gcc wrongly warns about y being used |
| 186 | // uninitialized unless we initialize it to something: |
| 187 | RealType x, y(0); |
| 188 | |
| 189 | x = boost::math::ibeta_inv(df1 / 2, df2 / 2, p, &y, Policy()); |
| 190 | |
| 191 | return df2 * x / (df1 * y); |
| 192 | } // quantile |
| 193 | |
| 194 | template <class RealType, class Policy> |
| 195 | inline RealType cdf(const complemented2_type<fisher_f_distribution<RealType, Policy>, RealType>& c) |
| 196 | { |
| 197 | static const char* function = "boost::math::cdf(fisher_f_distribution<%1%> const&, %1%)" ; |
| 198 | RealType df1 = c.dist.degrees_of_freedom1(); |
| 199 | RealType df2 = c.dist.degrees_of_freedom2(); |
| 200 | RealType x = c.param; |
| 201 | // Error check: |
| 202 | RealType error_result = 0; |
| 203 | if(false == detail::check_df( |
| 204 | function, df1, &error_result, Policy()) |
| 205 | && detail::check_df( |
| 206 | function, df2, &error_result, Policy())) |
| 207 | return error_result; |
| 208 | |
| 209 | if((x < 0) || !(boost::math::isfinite)(x)) |
| 210 | { |
| 211 | return policies::raise_domain_error<RealType>( |
| 212 | function, "Random Variable parameter was %1%, but must be > 0 !" , x, Policy()); |
| 213 | } |
| 214 | |
| 215 | RealType v1x = df1 * x; |
| 216 | // |
| 217 | // There are two equivalent formulas used here, the aim is |
| 218 | // to prevent the final argument to the incomplete beta |
| 219 | // from being too close to 1: for some values of df1 and df2 |
| 220 | // the rate of change can be arbitrarily large in this area, |
| 221 | // whilst the value we're passing will have lost information |
| 222 | // content as a result of being 0.999999something. Better |
| 223 | // to switch things around so we're passing 1-z instead. |
| 224 | // |
| 225 | return v1x > df2 |
| 226 | ? boost::math::ibeta(df2 / 2, df1 / 2, df2 / (df2 + v1x), Policy()) |
| 227 | : boost::math::ibetac(df1 / 2, df2 / 2, v1x / (df2 + v1x), Policy()); |
| 228 | } |
| 229 | |
| 230 | template <class RealType, class Policy> |
| 231 | inline RealType quantile(const complemented2_type<fisher_f_distribution<RealType, Policy>, RealType>& c) |
| 232 | { |
| 233 | static const char* function = "boost::math::quantile(fisher_f_distribution<%1%> const&, %1%)" ; |
| 234 | RealType df1 = c.dist.degrees_of_freedom1(); |
| 235 | RealType df2 = c.dist.degrees_of_freedom2(); |
| 236 | RealType p = c.param; |
| 237 | // Error check: |
| 238 | RealType error_result = 0; |
| 239 | if(false == (detail::check_df( |
| 240 | function, df1, &error_result, Policy()) |
| 241 | && detail::check_df( |
| 242 | function, df2, &error_result, Policy()) |
| 243 | && detail::check_probability( |
| 244 | function, p, &error_result, Policy()))) |
| 245 | return error_result; |
| 246 | |
| 247 | RealType x, y; |
| 248 | |
| 249 | x = boost::math::ibetac_inv(df1 / 2, df2 / 2, p, &y, Policy()); |
| 250 | |
| 251 | return df2 * x / (df1 * y); |
| 252 | } |
| 253 | |
| 254 | template <class RealType, class Policy> |
| 255 | inline RealType mean(const fisher_f_distribution<RealType, Policy>& dist) |
| 256 | { // Mean of F distribution = v. |
| 257 | static const char* function = "boost::math::mean(fisher_f_distribution<%1%> const&)" ; |
| 258 | RealType df1 = dist.degrees_of_freedom1(); |
| 259 | RealType df2 = dist.degrees_of_freedom2(); |
| 260 | // Error check: |
| 261 | RealType error_result = 0; |
| 262 | if(false == detail::check_df( |
| 263 | function, df1, &error_result, Policy()) |
| 264 | && detail::check_df( |
| 265 | function, df2, &error_result, Policy())) |
| 266 | return error_result; |
| 267 | if(df2 <= 2) |
| 268 | { |
| 269 | return policies::raise_domain_error<RealType>( |
| 270 | function, "Second degree of freedom was %1% but must be > 2 in order for the distribution to have a mean." , df2, Policy()); |
| 271 | } |
| 272 | return df2 / (df2 - 2); |
| 273 | } // mean |
| 274 | |
| 275 | template <class RealType, class Policy> |
| 276 | inline RealType variance(const fisher_f_distribution<RealType, Policy>& dist) |
| 277 | { // Variance of F distribution. |
| 278 | static const char* function = "boost::math::variance(fisher_f_distribution<%1%> const&)" ; |
| 279 | RealType df1 = dist.degrees_of_freedom1(); |
| 280 | RealType df2 = dist.degrees_of_freedom2(); |
| 281 | // Error check: |
| 282 | RealType error_result = 0; |
| 283 | if(false == detail::check_df( |
| 284 | function, df1, &error_result, Policy()) |
| 285 | && detail::check_df( |
| 286 | function, df2, &error_result, Policy())) |
| 287 | return error_result; |
| 288 | if(df2 <= 4) |
| 289 | { |
| 290 | return policies::raise_domain_error<RealType>( |
| 291 | function, "Second degree of freedom was %1% but must be > 4 in order for the distribution to have a valid variance." , df2, Policy()); |
| 292 | } |
| 293 | return 2 * df2 * df2 * (df1 + df2 - 2) / (df1 * (df2 - 2) * (df2 - 2) * (df2 - 4)); |
| 294 | } // variance |
| 295 | |
| 296 | template <class RealType, class Policy> |
| 297 | inline RealType mode(const fisher_f_distribution<RealType, Policy>& dist) |
| 298 | { |
| 299 | static const char* function = "boost::math::mode(fisher_f_distribution<%1%> const&)" ; |
| 300 | RealType df1 = dist.degrees_of_freedom1(); |
| 301 | RealType df2 = dist.degrees_of_freedom2(); |
| 302 | // Error check: |
| 303 | RealType error_result = 0; |
| 304 | if(false == detail::check_df( |
| 305 | function, df1, &error_result, Policy()) |
| 306 | && detail::check_df( |
| 307 | function, df2, &error_result, Policy())) |
| 308 | return error_result; |
| 309 | if(df1 <= 2) |
| 310 | { |
| 311 | return policies::raise_domain_error<RealType>( |
| 312 | function, "First degree of freedom was %1% but must be > 2 in order for the distribution to have a mode." , df1, Policy()); |
| 313 | } |
| 314 | return df2 * (df1 - 2) / (df1 * (df2 + 2)); |
| 315 | } |
| 316 | |
| 317 | //template <class RealType, class Policy> |
| 318 | //inline RealType median(const fisher_f_distribution<RealType, Policy>& dist) |
| 319 | //{ // Median of Fisher F distribution is not defined. |
| 320 | // return tools::domain_error<RealType>(BOOST_CURRENT_FUNCTION, "Median is not implemented, result is %1%!", std::numeric_limits<RealType>::quiet_NaN()); |
| 321 | // } // median |
| 322 | |
| 323 | // Now implemented via quantile(half) in derived accessors. |
| 324 | |
| 325 | template <class RealType, class Policy> |
| 326 | inline RealType skewness(const fisher_f_distribution<RealType, Policy>& dist) |
| 327 | { |
| 328 | static const char* function = "boost::math::skewness(fisher_f_distribution<%1%> const&)" ; |
| 329 | BOOST_MATH_STD_USING // ADL of std names |
| 330 | // See http://mathworld.wolfram.com/F-Distribution.html |
| 331 | RealType df1 = dist.degrees_of_freedom1(); |
| 332 | RealType df2 = dist.degrees_of_freedom2(); |
| 333 | // Error check: |
| 334 | RealType error_result = 0; |
| 335 | if(false == detail::check_df( |
| 336 | function, df1, &error_result, Policy()) |
| 337 | && detail::check_df( |
| 338 | function, df2, &error_result, Policy())) |
| 339 | return error_result; |
| 340 | if(df2 <= 6) |
| 341 | { |
| 342 | return policies::raise_domain_error<RealType>( |
| 343 | function, "Second degree of freedom was %1% but must be > 6 in order for the distribution to have a skewness." , df2, Policy()); |
| 344 | } |
| 345 | return 2 * (df2 + 2 * df1 - 2) * sqrt((2 * df2 - 8) / (df1 * (df2 + df1 - 2))) / (df2 - 6); |
| 346 | } |
| 347 | |
| 348 | template <class RealType, class Policy> |
| 349 | RealType kurtosis_excess(const fisher_f_distribution<RealType, Policy>& dist); |
| 350 | |
| 351 | template <class RealType, class Policy> |
| 352 | inline RealType kurtosis(const fisher_f_distribution<RealType, Policy>& dist) |
| 353 | { |
| 354 | return 3 + kurtosis_excess(dist); |
| 355 | } |
| 356 | |
| 357 | template <class RealType, class Policy> |
| 358 | inline RealType kurtosis_excess(const fisher_f_distribution<RealType, Policy>& dist) |
| 359 | { |
| 360 | static const char* function = "boost::math::kurtosis_excess(fisher_f_distribution<%1%> const&)" ; |
| 361 | // See http://mathworld.wolfram.com/F-Distribution.html |
| 362 | RealType df1 = dist.degrees_of_freedom1(); |
| 363 | RealType df2 = dist.degrees_of_freedom2(); |
| 364 | // Error check: |
| 365 | RealType error_result = 0; |
| 366 | if(false == detail::check_df( |
| 367 | function, df1, &error_result, Policy()) |
| 368 | && detail::check_df( |
| 369 | function, df2, &error_result, Policy())) |
| 370 | return error_result; |
| 371 | if(df2 <= 8) |
| 372 | { |
| 373 | return policies::raise_domain_error<RealType>( |
| 374 | function, "Second degree of freedom was %1% but must be > 8 in order for the distribution to have a kurtosis." , df2, Policy()); |
| 375 | } |
| 376 | RealType df2_2 = df2 * df2; |
| 377 | RealType df1_2 = df1 * df1; |
| 378 | RealType n = -16 + 20 * df2 - 8 * df2_2 + df2_2 * df2 + 44 * df1 - 32 * df2 * df1 + 5 * df2_2 * df1 - 22 * df1_2 + 5 * df2 * df1_2; |
| 379 | n *= 12; |
| 380 | RealType d = df1 * (df2 - 6) * (df2 - 8) * (df1 + df2 - 2); |
| 381 | return n / d; |
| 382 | } |
| 383 | |
| 384 | } // namespace math |
| 385 | } // namespace boost |
| 386 | |
| 387 | // This include must be at the end, *after* the accessors |
| 388 | // for this distribution have been defined, in order to |
| 389 | // keep compilers that support two-phase lookup happy. |
| 390 | #include <boost/math/distributions/detail/derived_accessors.hpp> |
| 391 | |
| 392 | #endif // BOOST_MATH_DISTRIBUTIONS_FISHER_F_HPP |
| 393 | |