001 /*
002 * Licensed to the Apache Software Foundation (ASF) under one or more
003 * contributor license agreements. See the NOTICE file distributed with
004 * this work for additional information regarding copyright ownership.
005 * The ASF licenses this file to You under the Apache License, Version 2.0
006 * (the "License"); you may not use this file except in compliance with
007 * the License. You may obtain a copy of the License at
008 *
009 * http://www.apache.org/licenses/LICENSE-2.0
010 *
011 * Unless required by applicable law or agreed to in writing, software
012 * distributed under the License is distributed on an "AS IS" BASIS,
013 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
014 * See the License for the specific language governing permissions and
015 * limitations under the License.
016 */
017 package org.apache.commons.math.stat.descriptive.moment;
018
019 import java.io.Serializable;
020
021 import org.apache.commons.math.stat.descriptive.AbstractStorelessUnivariateStatistic;
022 import org.apache.commons.math.stat.descriptive.WeightedEvaluation;
023 import org.apache.commons.math.stat.descriptive.summary.Sum;
024
025 /**
026 * <p>Computes the arithmetic mean of a set of values. Uses the definitional
027 * formula:</p>
028 * <p>
029 * mean = sum(x_i) / n
030 * </p>
031 * <p>where <code>n</code> is the number of observations.
032 * </p>
033 * <p>When {@link #increment(double)} is used to add data incrementally from a
034 * stream of (unstored) values, the value of the statistic that
035 * {@link #getResult()} returns is computed using the following recursive
036 * updating algorithm: </p>
037 * <ol>
038 * <li>Initialize <code>m = </code> the first value</li>
039 * <li>For each additional value, update using <br>
040 * <code>m = m + (new value - m) / (number of observations)</code></li>
041 * </ol>
042 * <p> If {@link #evaluate(double[])} is used to compute the mean of an array
043 * of stored values, a two-pass, corrected algorithm is used, starting with
044 * the definitional formula computed using the array of stored values and then
045 * correcting this by adding the mean deviation of the data values from the
046 * arithmetic mean. See, e.g. "Comparison of Several Algorithms for Computing
047 * Sample Means and Variances," Robert F. Ling, Journal of the American
048 * Statistical Association, Vol. 69, No. 348 (Dec., 1974), pp. 859-866. </p>
049 * <p>
050 * Returns <code>Double.NaN</code> if the dataset is empty.
051 * </p>
052 * <strong>Note that this implementation is not synchronized.</strong> If
053 * multiple threads access an instance of this class concurrently, and at least
054 * one of the threads invokes the <code>increment()</code> or
055 * <code>clear()</code> method, it must be synchronized externally.
056 *
057 * @version $Revision: 1006299 $ $Date: 2010-10-10 16:47:17 +0200 (dim. 10 oct. 2010) $
058 */
059 public class Mean extends AbstractStorelessUnivariateStatistic
060 implements Serializable, WeightedEvaluation {
061
062 /** Serializable version identifier */
063 private static final long serialVersionUID = -1296043746617791564L;
064
065 /** First moment on which this statistic is based. */
066 protected FirstMoment moment;
067
068 /**
069 * Determines whether or not this statistic can be incremented or cleared.
070 * <p>
071 * Statistics based on (constructed from) external moments cannot
072 * be incremented or cleared.</p>
073 */
074 protected boolean incMoment;
075
076 /** Constructs a Mean. */
077 public Mean() {
078 incMoment = true;
079 moment = new FirstMoment();
080 }
081
082 /**
083 * Constructs a Mean with an External Moment.
084 *
085 * @param m1 the moment
086 */
087 public Mean(final FirstMoment m1) {
088 this.moment = m1;
089 incMoment = false;
090 }
091
092 /**
093 * Copy constructor, creates a new {@code Mean} identical
094 * to the {@code original}
095 *
096 * @param original the {@code Mean} instance to copy
097 */
098 public Mean(Mean original) {
099 copy(original, this);
100 }
101
102 /**
103 * {@inheritDoc}
104 */
105 @Override
106 public void increment(final double d) {
107 if (incMoment) {
108 moment.increment(d);
109 }
110 }
111
112 /**
113 * {@inheritDoc}
114 */
115 @Override
116 public void clear() {
117 if (incMoment) {
118 moment.clear();
119 }
120 }
121
122 /**
123 * {@inheritDoc}
124 */
125 @Override
126 public double getResult() {
127 return moment.m1;
128 }
129
130 /**
131 * {@inheritDoc}
132 */
133 public long getN() {
134 return moment.getN();
135 }
136
137 /**
138 * Returns the arithmetic mean of the entries in the specified portion of
139 * the input array, or <code>Double.NaN</code> if the designated subarray
140 * is empty.
141 * <p>
142 * Throws <code>IllegalArgumentException</code> if the array is null.</p>
143 * <p>
144 * See {@link Mean} for details on the computing algorithm.</p>
145 *
146 * @param values the input array
147 * @param begin index of the first array element to include
148 * @param length the number of elements to include
149 * @return the mean of the values or Double.NaN if length = 0
150 * @throws IllegalArgumentException if the array is null or the array index
151 * parameters are not valid
152 */
153 @Override
154 public double evaluate(final double[] values,final int begin, final int length) {
155 if (test(values, begin, length)) {
156 Sum sum = new Sum();
157 double sampleSize = length;
158
159 // Compute initial estimate using definitional formula
160 double xbar = sum.evaluate(values, begin, length) / sampleSize;
161
162 // Compute correction factor in second pass
163 double correction = 0;
164 for (int i = begin; i < begin + length; i++) {
165 correction += values[i] - xbar;
166 }
167 return xbar + (correction/sampleSize);
168 }
169 return Double.NaN;
170 }
171
172 /**
173 * Returns the weighted arithmetic mean of the entries in the specified portion of
174 * the input array, or <code>Double.NaN</code> if the designated subarray
175 * is empty.
176 * <p>
177 * Throws <code>IllegalArgumentException</code> if either array is null.</p>
178 * <p>
179 * See {@link Mean} for details on the computing algorithm. The two-pass algorithm
180 * described above is used here, with weights applied in computing both the original
181 * estimate and the correction factor.</p>
182 * <p>
183 * Throws <code>IllegalArgumentException</code> if any of the following are true:
184 * <ul><li>the values array is null</li>
185 * <li>the weights array is null</li>
186 * <li>the weights array does not have the same length as the values array</li>
187 * <li>the weights array contains one or more infinite values</li>
188 * <li>the weights array contains one or more NaN values</li>
189 * <li>the weights array contains negative values</li>
190 * <li>the start and length arguments do not determine a valid array</li>
191 * </ul></p>
192 *
193 * @param values the input array
194 * @param weights the weights array
195 * @param begin index of the first array element to include
196 * @param length the number of elements to include
197 * @return the mean of the values or Double.NaN if length = 0
198 * @throws IllegalArgumentException if the parameters are not valid
199 * @since 2.1
200 */
201 public double evaluate(final double[] values, final double[] weights,
202 final int begin, final int length) {
203 if (test(values, weights, begin, length)) {
204 Sum sum = new Sum();
205
206 // Compute initial estimate using definitional formula
207 double sumw = sum.evaluate(weights,begin,length);
208 double xbarw = sum.evaluate(values, weights, begin, length) / sumw;
209
210 // Compute correction factor in second pass
211 double correction = 0;
212 for (int i = begin; i < begin + length; i++) {
213 correction += weights[i] * (values[i] - xbarw);
214 }
215 return xbarw + (correction/sumw);
216 }
217 return Double.NaN;
218 }
219
220 /**
221 * Returns the weighted arithmetic mean of the entries in the input array.
222 * <p>
223 * Throws <code>IllegalArgumentException</code> if either array is null.</p>
224 * <p>
225 * See {@link Mean} for details on the computing algorithm. The two-pass algorithm
226 * described above is used here, with weights applied in computing both the original
227 * estimate and the correction factor.</p>
228 * <p>
229 * Throws <code>IllegalArgumentException</code> if any of the following are true:
230 * <ul><li>the values array is null</li>
231 * <li>the weights array is null</li>
232 * <li>the weights array does not have the same length as the values array</li>
233 * <li>the weights array contains one or more infinite values</li>
234 * <li>the weights array contains one or more NaN values</li>
235 * <li>the weights array contains negative values</li>
236 * </ul></p>
237 *
238 * @param values the input array
239 * @param weights the weights array
240 * @return the mean of the values or Double.NaN if length = 0
241 * @throws IllegalArgumentException if the parameters are not valid
242 * @since 2.1
243 */
244 public double evaluate(final double[] values, final double[] weights) {
245 return evaluate(values, weights, 0, values.length);
246 }
247
248 /**
249 * {@inheritDoc}
250 */
251 @Override
252 public Mean copy() {
253 Mean result = new Mean();
254 copy(this, result);
255 return result;
256 }
257
258
259 /**
260 * Copies source to dest.
261 * <p>Neither source nor dest can be null.</p>
262 *
263 * @param source Mean to copy
264 * @param dest Mean to copy to
265 * @throws NullPointerException if either source or dest is null
266 */
267 public static void copy(Mean source, Mean dest) {
268 dest.setData(source.getDataRef());
269 dest.incMoment = source.incMoment;
270 dest.moment = source.moment.copy();
271 }
272 }