jidt/java/source/infodynamics/utils/AnalyticMeasurementDistribu...

147 lines
6.3 KiB
Java
Executable File

/*
* Java Information Dynamics Toolkit (JIDT)
* Copyright (C) 2012, Joseph T. Lizier
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
package infodynamics.utils;
/**
* Generic class to represent analytic distributions of info theoretic measurements under
* some null hypothesis of a relationship between the variables.
* This class is not designed to be creatable - it's children (which represent
* various concrete distributions) are those which should be created.
*
* @author Joseph Lizier (<a href="joseph.lizier at gmail.com">email</a>,
* <a href="http://lizier.me/joseph/">www</a>)
*/
public abstract class AnalyticMeasurementDistribution extends MeasurementDistribution {
/**
* Construct the instance
*
* @param actualValue observed value of the information-theoretic measure
* @param pValue p-value that surrogate measurements are greater than
* the observed value. (1 - CDF of null at the observed value)
*/
protected AnalyticMeasurementDistribution(double actualValue, double pValue) {
super(actualValue, pValue);
}
/**
* Compute the <b>analytic</b> p-value
* for the given estimate (i.e. the input argument, not the
* estimate produced by any particular observations supplied)
* under a null hypothesis that the source values of our
* samples had no relation to the destination value (possibly
* in the context of a conditional value).
*
* <p>See Section II.E "Statistical significance testing" of
* the JIDT paper below for a description of how this is done for MI,
* conditional MI and TE as per the references below.
* </p>
*
* <p><b>References:</b><br/>
* <ul>
* <li>J.T. Lizier, "JIDT: An information-theoretic
* toolkit for studying the dynamics of complex systems", 2014.</li>
* <li>Brillinger, <a href="http://www.stat.berkeley.edu/~brill/Papers/MIBJPS.pdf">
* "Some data analyses using mutual information"</a>,
* Brazilian Journal of Probability and Statistics, <b>18</b>, p. 163, (2004)</li>
* <li>Cheng et al., <a href="http://www.jds-online.com/file_download/112/JDS-369.pdf">
* "Data Information in Contingency Tables: A Fallacy of Hierarchical Loglinear Models"</a>,
* Journal of Data Science, <b>4</b>, p. 387 (2006).</li>
* <li>Geweke, <a href="http://dx.doi.org/10.1080/01621459.1982.10477803">
* "Measurement of Linear Dependence and Feedback between Multiple Time Series"</a>,
* Journal of the American Statistical Association, <b>77</b>, p. 304-313 (1982).</li>
* <li>Barnett and Bossomaier, <a href="http://arxiv.org/abs/1205.6339">
* "Transfer Entropy as a Log-likelihood Ratio"</a>,
* Physical Review Letters, <b>109</b>, p. 138105+ (2012).</li>
* </ul>
*
* @return p-value for the given channel measure score under this null hypothesis.
* @throws Exception
*/
public abstract double computePValueForGivenEstimate(double estimate);
/**
* Computes p-values corresponding to a set of estimates, each done via
* {@link #computePValueForGivenEstimate(double)}
*
* @param estimates array of estimates to return corresponding p-values for.
* @return
*/
public double[] computePValuesForGivenEstimates(double[] estimates) {
double[] pValues = new double[estimates.length];
for (int i = 0; i < estimates.length; i++) {
pValues[i] = computePValueForGivenEstimate(estimates[i]);
}
return estimates;
}
/**
* Compute the estimated observed measured value corresponding to
* a given p-value
* (derived from how the given estimates are <b>analytically</b> distributed
* under a null hypothesis that the source values of our
* samples had no relation to the destination value, possibly
* in the context of a conditional value).
*
* <p>See Section II.E "Statistical significance testing" of
* the JIDT paper below for a description of how this is done for MI,
* conditional MI and TE as per the references below.
* </p>
*
* <p><b>References:</b><br/>
* <ul>
* <li>J.T. Lizier, "JIDT: An information-theoretic
* toolkit for studying the dynamics of complex systems", 2014.</li>
* <li>Brillinger, <a href="http://www.stat.berkeley.edu/~brill/Papers/MIBJPS.pdf">
* "Some data analyses using mutual information"</a>,
* Brazilian Journal of Probability and Statistics, <b>18</b>, p. 163, (2004)</li>
* <li>Cheng et al., <a href="http://www.jds-online.com/file_download/112/JDS-369.pdf">
* "Data Information in Contingency Tables: A Fallacy of Hierarchical Loglinear Models"</a>,
* Journal of Data Science, <b>4</b>, p. 387 (2006).</li>
* <li>Geweke, <a href="http://dx.doi.org/10.1080/01621459.1982.10477803">
* "Measurement of Linear Dependence and Feedback between Multiple Time Series"</a>,
* Journal of the American Statistical Association, <b>77</b>, p. 304-313 (1982).</li>
* <li>Barnett and Bossomaier, <a href="http://arxiv.org/abs/1205.6339">
* "Transfer Entropy as a Log-likelihood Ratio"</a>,
* Physical Review Letters, <b>109</b>, p. 138105+ (2012).</li>
* </ul>
*
* @param pValue the sample p-value for the given channel measure score under this null hypothesis.
* @return the estimate of the channel measure score corresponding to
* the given p-value under this null hypothesis.
* @throws Exception
*/
public abstract double computeEstimateForGivenPValue(double pValue);
/**
* Computes estimates corresponding to a set of p-values, each done via
* {@link #computeEstimateForGivenPValue(double)}
*
* @param pValues array of p-values to return corresponding estimates for.
* @return
*/
public double[] computeEstimatesForGivenPValues(double[] pValues) {
double[] estimates = new double[pValues.length];
for (int i = 0; i < pValues.length; i++) {
estimates[i] = computeEstimateForGivenPValue(pValues[i]);
}
return estimates;
}
}