1 /*
2 * Licensed to the Apache Software Foundation (ASF) under one or more
3 * contributor license agreements. See the NOTICE file distributed with
4 * this work for additional information regarding copyright ownership.
5 * The ASF licenses this file to You under the Apache License, Version 2.0
6 * (the "License"); you may not use this file except in compliance with
7 * the License. You may obtain a copy of the License at
8 *
9 * http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 */
17 package org.apache.commons.rng.sampling.distribution;
18
19 import java.util.ArrayList;
20 import java.util.Arrays;
21 import java.util.List;
22 import org.apache.commons.math3.stat.inference.ChiSquareTest;
23 import org.junit.jupiter.api.Assertions;
24 import org.junit.jupiter.params.ParameterizedTest;
25 import org.junit.jupiter.params.provider.MethodSource;
26
27 /**
28 * Tests for random deviates generators.
29 */
30 class DiscreteSamplerParametricTest {
31 private static Iterable<DiscreteSamplerTestData> getSamplerTestData() {
32 return DiscreteSamplersList.list();
33 }
34
35 @ParameterizedTest
36 @MethodSource("getSamplerTestData")
37 void testSampling(DiscreteSamplerTestData data) {
38 final int sampleSize = 10000;
39 // Probabilities are normalised by the chi-square test
40 check(sampleSize,
41 data.getSampler(),
42 data.getPoints(),
43 data.getProbabilities());
44 }
45
46 /**
47 * Performs a chi-square test of homogeneity of the observed
48 * distribution with the expected distribution.
49 * An average failure rate higher than 5% causes the test case
50 * to fail.
51 *
52 * @param sampler Sampler.
53 * @param sampleSize Number of random values to generate.
54 * @param points Outcomes.
55 * @param expected Expected counts of the given outcomes.
56 */
57 private static void check(long sampleSize,
58 DiscreteSampler sampler,
59 int[] points,
60 double[] expected) {
61 final ChiSquareTest chiSquareTest = new ChiSquareTest();
62 final int numTests = 50;
63
64 // Run the tests.
65 int numFailures = 0;
66
67 final int numBins = points.length;
68 final long[] observed = new long[numBins];
69
70 // For storing chi2 larger than the critical value.
71 final List<Double> failedStat = new ArrayList<>();
72 try {
73 for (int i = 0; i < numTests; i++) {
74 Arrays.fill(observed, 0);
75 SAMPLE: for (long j = 0; j < sampleSize; j++) {
76 final int value = sampler.sample();
77
78 for (int k = 0; k < numBins; k++) {
79 if (value == points[k]) {
80 ++observed[k];
81 continue SAMPLE;
82 }
83 }
84 }
85
86 final double p = chiSquareTest.chiSquareTest(expected, observed);
87 if (p < 0.01) {
88 failedStat.add(p);
89 ++numFailures;
90 }
91 }
92 } catch (Exception e) {
93 // Should never happen.
94 throw new RuntimeException("Unexpected", e);
95 }
96
97 // The expected number of failed tests can be modelled as a Binomial distribution
98 // B(n, p) with n=50, p=0.01 (50 tests with a 1% significance level).
99 // The cumulative probability of the number of failed tests (X) is:
100 // x P(X>x)
101 // 1 0.0894
102 // 2 0.0138
103 // 3 0.0016
104
105 if (numFailures > 3) { // Test will fail with 0.16% probability
106 Assertions.fail(String.format(
107 "%s: Too many failures for sample size = %d " +
108 " (%d out of %d tests failed, chi2=%s",
109 sampler, sampleSize, numFailures, numTests,
110 Arrays.toString(failedStat.toArray(new Double[0]))));
111 }
112 }
113 }