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.statistics.descriptive; 18 19 /** 20 * Computes the kurtosis of the available values. The kurtosis is defined as: 21 * 22 * <p>\[ \operatorname{Kurt} = \operatorname{E}\left[ \left(\frac{X-\mu}{\sigma}\right)^4 \right] = \frac{\mu_4}{\sigma^4} \] 23 * 24 * <p>where \( \mu \) is the mean of \( X \), \( \sigma \) is the standard deviation of \( X \), 25 * \( \operatorname{E} \) represents the expectation operator, and \( \mu_4 \) is the fourth 26 * central moment. 27 * 28 * <p>The default implementation uses the following definition of the <em>sample kurtosis</em>: 29 * 30 * <p>\[ G_2 = \frac{k_4}{k_2^2} = \; 31 * \frac{n-1}{(n-2)\,(n-3)} \left[(n+1)\,\frac{m_4}{m_{2}^2} - 3\,(n-1) \right] \] 32 * 33 * <p>where \( k_4 \) is the unique symmetric unbiased estimator of the fourth cumulant, 34 * \( k_2 \) is the symmetric unbiased estimator of the second cumulant (i.e. the <em>sample variance</em>), 35 * \( m_4 \) is the fourth sample moment about the mean, 36 * \( m_2 \) is the second sample moment about the mean, 37 * \( \overline{x} \) is the sample mean, 38 * and \( n \) is the number of samples. 39 * 40 * <ul> 41 * <li>The result is {@code NaN} if less than 4 values are added. 42 * <li>The result is {@code NaN} if any of the values is {@code NaN} or infinite. 43 * <li>The result is {@code NaN} if the sum of the fourth deviations from the mean is infinite. 44 * </ul> 45 * 46 * <p>The default computation is for the adjusted Fisher–Pearson standardized moment coefficient 47 * \( G_2 \). If the {@link #setBiased(boolean) biased} option is enabled the following equation 48 * applies: 49 * 50 * <p>\[ g_2 = \frac{m_4}{m_2^2} - 3 = \frac{\tfrac{1}{n} \sum_{i=1}^n (x_i-\overline{x})^4} 51 * {\left[\tfrac{1}{n} \sum_{i=1}^n (x_i-\overline{x})^2 \right]^2} - 3 \] 52 * 53 * <p>In this case the computation only requires 2 values are added (i.e. the result is 54 * {@code NaN} if less than 2 values are added). 55 * 56 * <p>Note that the computation requires division by the second central moment \( m_2 \). 57 * If this is effectively zero then the result is {@code NaN}. This occurs when the value 58 * \( m_2 \) approaches the machine precision of the mean: \( m_2 \le (m_1 \times 10^{-15})^2 \). 59 * 60 * <p>The {@link #accept(double)} method uses a recursive updating algorithm. 61 * 62 * <p>The {@link #of(double...)} method uses a two-pass algorithm, starting with computation 63 * of the mean, and then computing the sum of deviations in a second pass. 64 * 65 * <p>Note that adding values using {@link #accept(double) accept} and then executing 66 * {@link #getAsDouble() getAsDouble} will 67 * sometimes give a different result than executing 68 * {@link #of(double...) of} with the full array of values. The former approach 69 * should only be used when the full array of values is not available. 70 * 71 * <p>Supports up to 2<sup>63</sup> (exclusive) observations. 72 * This implementation does not check for overflow of the count. 73 * 74 * <p>This class is designed to work with (though does not require) 75 * {@linkplain java.util.stream streams}. 76 * 77 * <p><strong>Note that this instance is not synchronized.</strong> If 78 * multiple threads access an instance of this class concurrently, and at least 79 * one of the threads invokes the {@link java.util.function.DoubleConsumer#accept(double) accept} or 80 * {@link StatisticAccumulator#combine(StatisticResult) combine} method, it must be synchronized externally. 81 * 82 * <p>However, it is safe to use {@link java.util.function.DoubleConsumer#accept(double) accept} 83 * and {@link StatisticAccumulator#combine(StatisticResult) combine} 84 * as {@code accumulator} and {@code combiner} functions of 85 * {@link java.util.stream.Collector Collector} on a parallel stream, 86 * because the parallel instance of {@link java.util.stream.Stream#collect Stream.collect()} 87 * provides the necessary partitioning, isolation, and merging of results for 88 * safe and efficient parallel execution. 89 * 90 * @see <a href="https://en.wikipedia.org/wiki/Kurtosis">Kurtosis (Wikipedia)</a> 91 * @since 1.1 92 */ 93 public final class Kurtosis implements DoubleStatistic, StatisticAccumulator<Kurtosis> { 94 /** 2, the length limit where the biased skewness is undefined. 95 * This limit effectively imposes the result m4 / m2^2 = 0 / 0 = NaN when 1 value 96 * has been added. Note that when more samples are added and the variance 97 * approaches zero the result is also returned as NaN. */ 98 private static final int LENGTH_TWO = 2; 99 /** 4, the length limit where the kurtosis is undefined. */ 100 private static final int LENGTH_FOUR = 4; 101 102 /** 103 * An instance of {@link SumOfFourthDeviations}, which is used to 104 * compute the kurtosis. 105 */ 106 private final SumOfFourthDeviations sq; 107 108 /** Flag to control if the statistic is biased, or should use a bias correction. */ 109 private boolean biased; 110 111 /** 112 * Create an instance. 113 */ 114 private Kurtosis() { 115 this(new SumOfFourthDeviations()); 116 } 117 118 /** 119 * Creates an instance with the sum of fourth deviations from the mean. 120 * 121 * @param sq Sum of fourth deviations. 122 */ 123 Kurtosis(SumOfFourthDeviations sq) { 124 this.sq = sq; 125 } 126 127 /** 128 * Creates an instance. 129 * 130 * <p>The initial result is {@code NaN}. 131 * 132 * @return {@code Kurtosis} instance. 133 */ 134 public static Kurtosis create() { 135 return new Kurtosis(); 136 } 137 138 /** 139 * Returns an instance populated using the input {@code values}. 140 * 141 * <p>Note: {@code Kurtosis} computed using {@link #accept(double) accept} may be 142 * different from this instance. 143 * 144 * @param values Values. 145 * @return {@code Kurtosis} instance. 146 */ 147 public static Kurtosis of(double... values) { 148 return new Kurtosis(SumOfFourthDeviations.of(values)); 149 } 150 151 /** 152 * Returns an instance populated using the input {@code values}. 153 * 154 * <p>Note: {@code Kurtosis} computed using {@link #accept(double) accept} may be 155 * different from this instance. 156 * 157 * @param values Values. 158 * @return {@code Kurtosis} instance. 159 */ 160 public static Kurtosis of(int... values) { 161 return new Kurtosis(SumOfFourthDeviations.of(values)); 162 } 163 164 /** 165 * Returns an instance populated using the input {@code values}. 166 * 167 * <p>Note: {@code Kurtosis} computed using {@link #accept(double) accept} may be 168 * different from this instance. 169 * 170 * @param values Values. 171 * @return {@code Kurtosis} instance. 172 */ 173 public static Kurtosis of(long... values) { 174 return new Kurtosis(SumOfFourthDeviations.of(values)); 175 } 176 177 /** 178 * Updates the state of the statistic to reflect the addition of {@code value}. 179 * 180 * @param value Value. 181 */ 182 @Override 183 public void accept(double value) { 184 sq.accept(value); 185 } 186 187 /** 188 * Gets the kurtosis of all input values. 189 * 190 * <p>When fewer than 4 values have been added, the result is {@code NaN}. 191 * 192 * @return kurtosis of all values. 193 */ 194 @Override 195 public double getAsDouble() { 196 // This method checks the sum of squared or fourth deviations is finite 197 // to provide a consistent NaN when the computation is not possible. 198 199 if (sq.n < (biased ? LENGTH_TWO : LENGTH_FOUR)) { 200 return Double.NaN; 201 } 202 final double x2 = sq.getSumOfSquaredDeviations(); 203 if (!Double.isFinite(x2)) { 204 return Double.NaN; 205 } 206 final double x4 = sq.getSumOfFourthDeviations(); 207 if (!Double.isFinite(x4)) { 208 return Double.NaN; 209 } 210 // Avoid a divide by zero; for a negligible variance return NaN. 211 // Note: Commons Math returns zero if variance is < 1e-19. 212 final double m2 = x2 / sq.n; 213 if (Statistics.zeroVariance(sq.getFirstMoment(), m2)) { 214 return Double.NaN; 215 } 216 final double m4 = x4 / sq.n; 217 if (biased) { 218 return m4 / (m2 * m2) - 3; 219 } 220 final double n = sq.n; 221 return ((n * n - 1) * m4 / (m2 * m2) - 3 * (n - 1) * (n - 1)) / ((n - 2) * (n - 3)); 222 } 223 224 @Override 225 public Kurtosis combine(Kurtosis other) { 226 sq.combine(other.sq); 227 return this; 228 } 229 230 /** 231 * Sets the value of the biased flag. The default value is {@code false}. 232 * See {@link Kurtosis} for details on the computing algorithm. 233 * 234 * <p>This flag only controls the final computation of the statistic. The value of this flag 235 * will not affect compatibility between instances during a {@link #combine(Kurtosis) combine} 236 * operation. 237 * 238 * @param v Value. 239 * @return {@code this} instance 240 */ 241 public Kurtosis setBiased(boolean v) { 242 biased = v; 243 return this; 244 } 245 }