| 1 | /* -*- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */ |
| 2 | |
| 3 | /* |
| 4 | Copyright (C) 2012 Liquidnet Holdings, Inc. |
| 5 | |
| 6 | This file is part of QuantLib, a free-software/open-source library |
| 7 | for financial quantitative analysts and developers - http://quantlib.org/ |
| 8 | |
| 9 | QuantLib is free software: you can redistribute it and/or modify it |
| 10 | under the terms of the QuantLib license. You should have received a |
| 11 | copy of the license along with this program; if not, please email |
| 12 | <quantlib-dev@lists.sf.net>. The license is also available online at |
| 13 | <http://quantlib.org/license.shtml>. |
| 14 | |
| 15 | This program is distributed in the hope that it will be useful, but WITHOUT |
| 16 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS |
| 17 | FOR A PARTICULAR PURPOSE. See the license for more details. |
| 18 | */ |
| 19 | |
| 20 | #include "garch.hpp" |
| 21 | #include "utilities.hpp" |
| 22 | #include <ql/models/volatility/garch.hpp> |
| 23 | #include <ql/time/calendars/target.hpp> |
| 24 | #include <ql/math/optimization/levenbergmarquardt.hpp> |
| 25 | #include <ql/math/randomnumbers/inversecumulativerng.hpp> |
| 26 | #include <ql/math/randomnumbers/mt19937uniformrng.hpp> |
| 27 | #include <ql/math/distributions/normaldistribution.hpp> |
| 28 | |
| 29 | using namespace QuantLib; |
| 30 | using boost::unit_test_framework::test_suite; |
| 31 | |
| 32 | namespace garch_test { |
| 33 | |
| 34 | class DummyOptimizationMethod : public OptimizationMethod { |
| 35 | public: |
| 36 | EndCriteria::Type minimize(Problem& P, const EndCriteria& endCriteria) override { |
| 37 | P.setFunctionValue(P.value(x: P.currentValue())); |
| 38 | return EndCriteria::None; |
| 39 | } |
| 40 | }; |
| 41 | |
| 42 | struct Results { |
| 43 | Real alpha; |
| 44 | Real beta; |
| 45 | Real omega; |
| 46 | Real logLikelihood; |
| 47 | }; |
| 48 | |
| 49 | typedef InverseCumulativeRng<MersenneTwisterUniformRng, |
| 50 | InverseCumulativeNormal> |
| 51 | GaussianGenerator; |
| 52 | } |
| 53 | |
| 54 | #define CHECK(results, garch, member) \ |
| 55 | if (std::fabs(results.member - garch.member()) > 1.0e-6) { \ |
| 56 | BOOST_ERROR("Failed to reproduce expected " #member \ |
| 57 | << "\n calculated: " << garch.member() \ |
| 58 | << "\n expected: " << results.member); \ |
| 59 | } |
| 60 | |
| 61 | void GARCHTest::testCalibration() { |
| 62 | |
| 63 | BOOST_TEST_MESSAGE("Testing GARCH model calibration..." ); |
| 64 | |
| 65 | using namespace garch_test; |
| 66 | |
| 67 | Date start(7, July, 1962), d = start; |
| 68 | TimeSeries<Volatility> ts; |
| 69 | Garch11 garch(0.2, 0.3, 0.4); |
| 70 | GaussianGenerator rng(MersenneTwisterUniformRng(48)); |
| 71 | |
| 72 | Volatility r = 0.0, v = 0.0; |
| 73 | for (std::size_t i = 0; i < 50000; ++i, d += 1) { |
| 74 | v = garch.forecast(r, sigma2: v); |
| 75 | r = rng.next().value * std::sqrt(x: v); |
| 76 | ts[d] = r; |
| 77 | } |
| 78 | |
| 79 | // Default calibration; works fine in most cases |
| 80 | Garch11 cgarch1(ts); |
| 81 | |
| 82 | Results calibrated = { .alpha: 0.207592, .beta: 0.281979, .omega: 0.204647, .logLikelihood: -0.0217413 }; |
| 83 | |
| 84 | CHECK(calibrated, cgarch1, alpha); |
| 85 | CHECK(calibrated, cgarch1, beta); |
| 86 | CHECK(calibrated, cgarch1, omega); |
| 87 | CHECK(calibrated, cgarch1, logLikelihood); |
| 88 | |
| 89 | // Type 1 initial guess - no further optimization |
| 90 | Garch11 cgarch2(ts, Garch11::MomentMatchingGuess); |
| 91 | DummyOptimizationMethod m; |
| 92 | cgarch2.calibrate(quoteSeries: ts, method&: m, endCriteria: EndCriteria (3, 2, 0.0, 0.0, 0.0)); |
| 93 | Results expected1 = { .alpha: 0.265749, .beta: 0.156956, .omega: 0.230964, .logLikelihood: -0.0227179 }; |
| 94 | |
| 95 | CHECK(expected1, cgarch2, alpha); |
| 96 | CHECK(expected1, cgarch2, beta); |
| 97 | CHECK(expected1, cgarch2, omega); |
| 98 | CHECK(expected1, cgarch2, logLikelihood); |
| 99 | |
| 100 | // Optimization from this initial guess |
| 101 | cgarch2.calibrate(quoteSeries: ts); |
| 102 | |
| 103 | CHECK(calibrated, cgarch2, alpha); |
| 104 | CHECK(calibrated, cgarch2, beta); |
| 105 | CHECK(calibrated, cgarch2, omega); |
| 106 | CHECK(calibrated, cgarch2, logLikelihood); |
| 107 | |
| 108 | // Type 2 initial guess - no further optimization |
| 109 | Garch11 cgarch3(ts, Garch11::GammaGuess); |
| 110 | cgarch3.calibrate(quoteSeries: ts, method&: m, endCriteria: EndCriteria (3, 2, 0.0, 0.0, 0.0)); |
| 111 | Results expected2 = { .alpha: 0.269896, .beta: 0.211373, .omega: 0.207534, .logLikelihood: -0.022798 }; |
| 112 | |
| 113 | CHECK(expected2, cgarch3, alpha); |
| 114 | CHECK(expected2, cgarch3, beta); |
| 115 | CHECK(expected2, cgarch3, omega); |
| 116 | CHECK(expected2, cgarch3, logLikelihood); |
| 117 | |
| 118 | // Optimization from this initial guess |
| 119 | cgarch3.calibrate(quoteSeries: ts); |
| 120 | |
| 121 | CHECK(calibrated, cgarch3, alpha); |
| 122 | CHECK(calibrated, cgarch3, beta); |
| 123 | CHECK(calibrated, cgarch3, omega); |
| 124 | CHECK(calibrated, cgarch3, logLikelihood); |
| 125 | |
| 126 | // Double optimization using type 1 and 2 initial guesses |
| 127 | Garch11 cgarch4(ts, Garch11::DoubleOptimization); |
| 128 | cgarch4.calibrate(quoteSeries: ts); |
| 129 | |
| 130 | CHECK(calibrated, cgarch4, alpha); |
| 131 | CHECK(calibrated, cgarch4, beta); |
| 132 | CHECK(calibrated, cgarch4, omega); |
| 133 | CHECK(calibrated, cgarch4, logLikelihood); |
| 134 | |
| 135 | // Alternative, gradient based optimization - usually gives worse |
| 136 | // results than simplex |
| 137 | LevenbergMarquardt lm; |
| 138 | cgarch4.calibrate(quoteSeries: ts, method&: lm, endCriteria: EndCriteria (100000, 500, 1e-8, 1e-8, 1e-8)); |
| 139 | Results expected3 = { .alpha: 0.265196, .beta: 0.277364, .omega: 0.678812, .logLikelihood: -0.216313 }; |
| 140 | |
| 141 | CHECK(expected3, cgarch4, alpha); |
| 142 | CHECK(expected3, cgarch4, beta); |
| 143 | CHECK(expected3, cgarch4, omega); |
| 144 | CHECK(expected3, cgarch4, logLikelihood); |
| 145 | } |
| 146 | |
| 147 | namespace garch_test { |
| 148 | |
| 149 | static Real expected_calc[] = { |
| 150 | 0.452769, 0.513323, 0.530141, 0.5350841, 0.536558, |
| 151 | 0.536999, 0.537132, 0.537171, 0.537183, 0.537187 |
| 152 | }; |
| 153 | |
| 154 | void check_ts(const std::pair<Date, Volatility> &x) { |
| 155 | if (x.first.serialNumber() < 22835 || x.first.serialNumber() > 22844) { |
| 156 | BOOST_ERROR("Failed to reproduce calculated GARCH time: " |
| 157 | << "\n calculated: " << x.first.serialNumber() |
| 158 | << "\n expected: [22835, 22844]" ); |
| 159 | } |
| 160 | Real error = |
| 161 | std::fabs(x: x.second - expected_calc[x.first.serialNumber()-22835]); |
| 162 | if (error > 1.0e-6) { |
| 163 | BOOST_ERROR("Failed to reproduce calculated GARCH value at " |
| 164 | << x.first.serialNumber() << ": " |
| 165 | << "\n calculated: " << x.second |
| 166 | << "\n expected: " |
| 167 | << expected_calc[x.first.serialNumber()-22835]); |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | } |
| 172 | |
| 173 | void GARCHTest::testCalculation() { |
| 174 | BOOST_TEST_MESSAGE("Testing GARCH model calculation..." ); |
| 175 | |
| 176 | using namespace garch_test; |
| 177 | |
| 178 | Date d(7, July, 1962); |
| 179 | TimeSeries<Volatility> ts; |
| 180 | Garch11 garch(0.2, 0.3, 0.4); |
| 181 | |
| 182 | Volatility r = 0.1; |
| 183 | for (std::size_t i = 0; i < 10; ++i, d += 1) { |
| 184 | ts[d] = r; |
| 185 | } |
| 186 | |
| 187 | TimeSeries<Volatility> tsout = garch.calculate(quoteSeries: ts); |
| 188 | std::for_each(first: tsout.cbegin(), last: tsout.cend(), f: check_ts); |
| 189 | } |
| 190 | |
| 191 | test_suite* GARCHTest::suite() { |
| 192 | auto* suite = BOOST_TEST_SUITE("GARCH model tests" ); |
| 193 | suite->add(QUANTLIB_TEST_CASE(&GARCHTest::testCalibration)); |
| 194 | suite->add(QUANTLIB_TEST_CASE(&GARCHTest::testCalculation)); |
| 195 | return suite; |
| 196 | } |
| 197 | |