icb-stats API icb-stats-31 (20120309094430)

edu.cornell.med.icb.stat
Class MinMaxCalculator

java.lang.Object
  extended by edu.cornell.med.icb.stat.MinMaxCalculator

public class MinMaxCalculator
extends Object

Calculates the Min/Max statistics for a feature. Each feature must be given as a list of value and associated class label. For each feature, we determine the minimum and maximum feature values in the positive and in the negative class. We refer to these values as minL0, minL1, maxL0 and maxL1. The Min/Max statistics is defined as (Math.abs(maxL1 - minL0) / Math.abs(minL1 - maxL0)). This class also supports averaging the k smallest and k largest observations of the feature values to estimate minL0, minL1, maxL0 and maxL1 in a more robust manner. The statistics is larger when the separation of minimum and maximum values of a feature among classes suggest that the feature may be predictive in a subset of the samples observed.

Author:
Fabien Campagne Date: Apr 1, 2008 Time: 4:13:33 PM

Field Summary
static Log LOG
          The logger to use.
 
Constructor Summary
MinMaxCalculator()
           
 
Method Summary
static double compute(double[] values, double[] labels)
           
static double compute(double[] values, double[] labels, int k)
          Compute the min/max statistics.
static void main(String[] args)
           
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Field Detail

LOG

public static final Log LOG
The logger to use.

Constructor Detail

MinMaxCalculator

public MinMaxCalculator()
Method Detail

compute

public static double compute(double[] values,
                             double[] labels)

compute

public static double compute(double[] values,
                             double[] labels,
                             int k)
Compute the min/max statistics.

Parameters:
values -
labels -
k -
Returns:

main

public static void main(String[] args)

icb-stats API icb-stats-31 (20120309094430)

Copyright © 2008-2012 Institute for Computational Biomedicine, All Rights Reserved.