weka.core.stats

Class NumericStats

java.lang.Object

weka.core.stats.Stats

weka.core.stats.NumericStats

All Implemented Interfaces:

java.io.Serializable

public class NumericStats
extends Stats
implements java.io.Serializable

Class for computing numeric stats

Version:

$Revision: 13233 $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
static double	Q_COMPRESSION Default compression for TDigest quantile estimators

Constructor Summary

Constructors

Constructor and Description
NumericStats(java.lang.String attributeName) Constructs a new NumericStats
NumericStats(java.lang.String attributeName, double quantileCompression) Construct a new NumericStats

Method Summary

Modifier and Type	Method and Description
static NumericStats	attributeToStats(weka.core.Attribute a) Convert a summary meta attribute into a NumericStats object (does not recover the internal TDigest quantile estimator)
void	computeDerived() Compute the derived statistics
void	computeQuartilesAndHistogram() Computes derived stats and computes quartiles and histogram data from our quantile estimator.
void	deSerializeCurrentQuantileEstimator() Decode the current TDigest quatile estimator
java.util.List<java.lang.String>	getHistogramBinLabels() Get the histogram labels
java.util.List<java.lang.Double>	getHistogramFrequencies() Get the histogram bin frequencies
TDigest	getQuantileEstimator() Get the quantile estimator in use (if any)
double[]	getStats() Return the array of statistics
weka.core.Attribute	makeAttribute() Makes a Attribute that encapsulates the meta data
void	serializeCurrentQuantileEstimator() Serialize the current TDigest quantile estimator
void	setCompression(double compression) Set the compression level
void	setHistogramData(java.util.List<java.lang.String> labs, java.util.List<java.lang.Double> freqs) Set histogram data for this numeric stats
void	setQuantileEstimator(TDigest estimator) Set the quantile estimator to use
void	setStats(double[] stats) Sets the array of statistics.
void	update(double value, double weight, boolean treatZeroAsMissing, boolean updateQuantiles) Update the incremental aggregateable portions of this NumericStats with the supplied value

Methods inherited from class weka.core.stats.Stats

getName

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

Q_COMPRESSION

public static final double Q_COMPRESSION

Default compression for TDigest quantile estimators

See Also:

Constant Field Values

Constructor Detail

NumericStats

public NumericStats(java.lang.String attributeName)

Constructs a new NumericStats

Parameters:

attributeName - the name of the attribute that these statistics are for

NumericStats

public NumericStats(java.lang.String attributeName,
                    double quantileCompression)

Construct a new NumericStats

Parameters:

attributeName - the name of the attribute that these statistics are for

quantileCompression - the degree of compression for quantile estimation (bigger = less compression)

Method Detail

update

public void update(double value,
                   double weight,
                   boolean treatZeroAsMissing,
                   boolean updateQuantiles)

Update the incremental aggregateable portions of this NumericStats with the supplied value

Parameters:

value - the value to update with

weight - the weight to use

treatZeroAsMissing - true if zeros count as missing

updateQuantiles - true if we should update our quantile estimator

getStats

public double[] getStats()

Return the array of statistics

Returns:

the array of statistics

setStats

public void setStats(double[] stats)

Sets the array of statistics. Does not check to see if the supplied array is of the correct length.

Parameters:

stats - the stats array to use

getQuantileEstimator

public TDigest getQuantileEstimator()

Get the quantile estimator in use (if any)

Returns:

the quantile estmator

setQuantileEstimator

public void setQuantileEstimator(TDigest estimator)

Set the quantile estimator to use

Parameters:

estimator - the estimator to use

setCompression

public void setCompression(double compression)

Set the compression level

Parameters:

compression - compression level

serializeCurrentQuantileEstimator

public void serializeCurrentQuantileEstimator()

Serialize the current TDigest quantile estimator

deSerializeCurrentQuantileEstimator

public void deSerializeCurrentQuantileEstimator()

Decode the current TDigest quatile estimator

setHistogramData

public void setHistogramData(java.util.List<java.lang.String> labs,
                             java.util.List<java.lang.Double> freqs)

Set histogram data for this numeric stats

Parameters:

labs - bin labels

freqs - bin frequencies

getHistogramBinLabels

public java.util.List<java.lang.String> getHistogramBinLabels()

Get the histogram labels

Returns:

the list of histogram labels or null if not set

getHistogramFrequencies

public java.util.List<java.lang.Double> getHistogramFrequencies()

Get the histogram bin frequencies

Returns:

the list of histogram bin frequencies or null if not set

makeAttribute

public weka.core.Attribute makeAttribute()

Description copied from class: Stats

Makes a Attribute that encapsulates the meta data

Specified by:

makeAttribute in class Stats

Returns:

an Attribute that encapsulates the meta data

attributeToStats

public static NumericStats attributeToStats(weka.core.Attribute a)
                                     throws java.lang.IllegalArgumentException

Convert a summary meta attribute into a NumericStats object (does not recover the internal TDigest quantile estimator)

Parameters:

a - the summary meta attribute to convert

Returns:

a NumericStats instance

Throws:

java.lang.IllegalArgumentException - if a problem occurs

computeDerived

public void computeDerived()

Compute the derived statistics

computeQuartilesAndHistogram

public void computeQuartilesAndHistogram()

Computes derived stats and computes quartiles and histogram data from our quantile estimator. If quartiles are not being estimated then the result is just to call computeDerived().