// (C) Copyright Eric Niebler 2005. // Use, modification and distribution are subject to the // Boost Software License, Version 1.0. (See accompanying file // LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) // Test case for pot_quantile.hpp (weighted feature) #define BOOST_NUMERIC_FUNCTIONAL_STD_COMPLEX_SUPPORT #define BOOST_NUMERIC_FUNCTIONAL_STD_VALARRAY_SUPPORT #define BOOST_NUMERIC_FUNCTIONAL_STD_VECTOR_SUPPORT #include #include #include #include #include using namespace boost; using namespace unit_test; using namespace boost::accumulators; /////////////////////////////////////////////////////////////////////////////// // test_stat // void test_stat() { // tolerance in % double epsilon = 1.; double mu1, mu2, l; mu1 = 1.; mu2 = -1.; l = 0.5; // two random number generators boost::lagged_fibonacci607 rng; boost::normal_distribution<> mean_sigma1(mu1,1); boost::normal_distribution<> mean_sigma2(mu2,1); boost::exponential_distribution<> lambda(l); boost::variate_generator > normal1(rng, mean_sigma1); boost::variate_generator > normal2(rng, mean_sigma2); boost::variate_generator > exponential(rng, lambda); accumulator_set(with_threshold_value)>, double > acc1( pot_threshold_value = 3. ); accumulator_set(with_threshold_probability)>, double > acc2( right_tail_cache_size = 10000 , pot_threshold_probability = 0.99 ); accumulator_set(with_threshold_value)>, double > acc3( pot_threshold_value = -3. ); accumulator_set(with_threshold_probability)>, double > acc4( left_tail_cache_size = 10000 , pot_threshold_probability = 0.01 ); accumulator_set(with_threshold_value)>, double > acc5( pot_threshold_value = 5. ); accumulator_set(with_threshold_probability)>, double > acc6( right_tail_cache_size = 10000 , pot_threshold_probability = 0.995 ); for (std::size_t i = 0; i < 100000; ++i) { double sample1 = normal1(); double sample2 = normal2(); acc1(sample1, weight = std::exp(-mu1 * (sample1 - 0.5 * mu1))); acc2(sample1, weight = std::exp(-mu1 * (sample1 - 0.5 * mu1))); acc3(sample2, weight = std::exp(-mu2 * (sample2 - 0.5 * mu2))); acc4(sample2, weight = std::exp(-mu2 * (sample2 - 0.5 * mu2))); } for (std::size_t i = 0; i < 100000; ++i) { double sample = exponential(); acc5(sample, weight = 1./l * std::exp(-sample * (1. - l))); acc6(sample, weight = 1./l * std::exp(-sample * (1. - l))); } BOOST_CHECK_CLOSE( quantile(acc1, quantile_probability = 0.999), 3.090232, epsilon ); BOOST_CHECK_CLOSE( quantile(acc2, quantile_probability = 0.999), 3.090232, epsilon ); BOOST_CHECK_CLOSE( quantile(acc3, quantile_probability = 0.001), -3.090232, epsilon ); BOOST_CHECK_CLOSE( quantile(acc4, quantile_probability = 0.001), -3.090232, epsilon ); BOOST_CHECK_CLOSE( quantile(acc5, quantile_probability = 0.999), 6.908, epsilon ); BOOST_CHECK_CLOSE( quantile(acc6, quantile_probability = 0.999), 6.908, epsilon ); } /////////////////////////////////////////////////////////////////////////////// // init_unit_test_suite // test_suite* init_unit_test_suite( int argc, char* argv[] ) { test_suite *test = BOOST_TEST_SUITE("weighted_pot_quantile test"); test->add(BOOST_TEST_CASE(&test_stat)); return test; }