Package etomica.data

Class AccumulatorAverage

java.lang.Object

etomica.data.DataProcessor

etomica.data.DataProcessorForked

etomica.data.DataAccumulator

etomica.data.AccumulatorAverage

All Implemented Interfaces:

DataPipe, DataPipeForked, IDataSink, IDataSource

Direct Known Subclasses:

AccumulatorAverageCollapsing, AccumulatorAverageFixed, AccumulatorAverageFixedOutputFile

public abstract class AccumulatorAverage
extends DataAccumulator

Accumulator that keeps statistics for averaging and error analysis. The statistics are incremented with every call to addData. Statistics kept are

• the value last given to addData
• the mean, or average
• the statistical error (68% confidence limits), obtained by standard error analysis of block averages
• the standard deviation
• the most recent block average
• the correlation between adjacent blocks (covariance divided by variance)

Each statistic is recorded by a Data instance of the same type as the incoming data. The output Data is a DataGroup formed from these Data instances (in the order given above).

A block is a subset of the input data, formed by grouping (for example) 100 or 1000 successive contributions to addData. Averages for each block are considered to be independent of other block averages, and the confidence interval for the overall average is obtained as the standard error of the mean of these block averages.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	AccumulatorAverage.StatType	Enumerated type that can be used to indicate the statistic to be taken from the accumulator (e.g., average, error, current value, etc.).

Field Summary

Fields

Modifier and Type	Field	Description
protected IData	average
static AccumulatorAverage.StatType	AVERAGE
static AccumulatorAverage.StatType	BLOCK_CORRELATION
protected IData	blockCorrelation
protected long	blockCountDown
protected long	blockSize
protected long	count
protected DataGroup	dataGroup
protected boolean	doIncludeACInError
protected IData	error
static AccumulatorAverage.StatType	ERROR
static AccumulatorAverage.StatType	MOST_RECENT
protected IData	mostRecent
static AccumulatorAverage.StatType	STANDARD_DEVIATION
protected IData	standardDeviation

Fields inherited from class etomica.data.DataAccumulator

active, hasUnpushedData, pushInterval, putCount

Fields inherited from class etomica.data.DataProcessor

dataInfo, dataSink, tag

Constructor Summary

Constructors

Constructor	Description
AccumulatorAverage()	Default constructor sets block size to 1000 and sets the interval for pushing the output data (pushInterval) to 100.
AccumulatorAverage(long blockSize)	Constructs with interval for pushing the output data set to 100.

Method Summary

Modifier and Type	Method	Description
void	addDataSink(IDataSink newDataSink, AccumulatorAverage.StatType[] types)	Adds a new DataSink that will receive a specific subset of the statistics generated by this accumulator.
boolean	doIncludeACInError()
long	getBlockCount()
long	getBlockSize()
IData	getData(AccumulatorAverage.StatType stat)	Returns a specific element from the DataGroup returned by getData().
long	getSampleCount()
DataTag	getTag(AccumulatorAverage.StatType statType)
IDataInfo	processDataInfo(IDataInfo incomingDataInfo)	Prepares the accumulator for input data.
void	reset()	Resets all sums to zero.
void	setBlockSize(long newBlockSize)	Sets the size of the block used to group data for error analysis.
void	setIncludeACInError(boolean newDoIncludeACInError)	Include correction to the uncertainty from block correlation function as prescribed in Kolafa, Jiri(1986) 'Autocorrelations and subseries averages in Monte Carlo Simulations', Molecular Physics (59) 1035
static AccumulatorAverage.StatType[]	statChoices()

Methods inherited from class etomica.data.DataAccumulator

addData, getData, getPushInterval, isActive, processData, setActive, setPushInterval, toString

Methods inherited from interface etomica.data.DataPipe

getTag

Methods inherited from class etomica.data.DataProcessor

getDataInfo, getTag, putData, putDataInfo

Methods inherited from class etomica.data.DataProcessorForked

addDataSink, getDataSink, getDataSinks, removeDataSink, setDataSink, setDataSinks

Methods inherited from interface etomica.data.IDataSink

putData, putDataInfo

Methods inherited from interface etomica.data.IDataSource

getDataInfo, getTag

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

MOST_RECENT

public static final AccumulatorAverage.StatType MOST_RECENT

AVERAGE

public static final AccumulatorAverage.StatType AVERAGE

ERROR

public static final AccumulatorAverage.StatType ERROR

STANDARD_DEVIATION

public static final AccumulatorAverage.StatType STANDARD_DEVIATION

BLOCK_CORRELATION

public static final AccumulatorAverage.StatType BLOCK_CORRELATION

mostRecent

protected IData mostRecent

average

protected IData average

error

protected IData error

standardDeviation

protected IData standardDeviation

blockCorrelation

protected IData blockCorrelation

dataGroup

protected DataGroup dataGroup

count

protected long count

blockCountDown

protected long blockCountDown

blockSize

protected long blockSize

doIncludeACInError

protected boolean doIncludeACInError

Constructor Detail

AccumulatorAverage

public AccumulatorAverage()

Default constructor sets block size to 1000 and sets the interval for pushing the output data (pushInterval) to 100.

AccumulatorAverage

public AccumulatorAverage(long blockSize)

Constructs with interval for pushing the output data set to 100.

Parameters:

blockSize - size of the block.

Method Detail

statChoices

public static AccumulatorAverage.StatType[] statChoices()

Returns:

array of StatTypes identifying the statistics given in the DataGroup returned by getData().

getTag

public DataTag getTag(AccumulatorAverage.StatType statType)

Parameters:

statType - the statistic for which the DataTag is requested.

Returns:

the DataTag that identifies the specified statistic.

getData

public IData getData(AccumulatorAverage.StatType stat)

Returns a specific element from the DataGroup returned by getData(). Calling getData and extracting the data can be more efficient, while this method can be more convenient.

Returns:

an IData object corresponding to the requested StatType.

reset

public void reset()

Resets all sums to zero. All statistics are cleared.

Specified by:

reset in class DataAccumulator

getBlockSize

public long getBlockSize()

Returns:

the current value of the block size.

setBlockSize

public void setBlockSize(long newBlockSize)

Sets the size of the block used to group data for error analysis. Resets statistics accumulated so far.

Parameters:

newBlockSize - new block size.

processDataInfo

public IDataInfo processDataInfo(IDataInfo incomingDataInfo)

Prepares the accumulator for input data. Discards any previous contributions to statistics.

Specified by:

processDataInfo in class DataProcessor

Parameters:

incomingDataInfo - the DataInfo instance for the data that will be given to addData

Returns:

dataInfo object for the output of this accumulator.

addDataSink

public void addDataSink(IDataSink newDataSink,
                        AccumulatorAverage.StatType[] types)

Adds a new DataSink that will receive a specific subset of the statistics generated by this accumulator. The DataSink will be given a DataGroup that has only the specified statistics in it. The output of this accumulator is passed through a DataGroupFilter that removes the unwanted Data.

Parameters:

newDataSink - the new DataSink

types - array indicating the statistics to be included in the DataGroup sent to the sink.

getBlockCount

public long getBlockCount()

Returns:

the number of blocks that have been accumulated since construction or the last call to reset.

getSampleCount

public long getSampleCount()

Returns:

the number of samples that have been accumulated since construction or the last call to reset.

doIncludeACInError

public boolean doIncludeACInError()

Returns:

Flag indicating if auto-correlation is accounted for while calculating the uncertainty.

See Also:

setIncludeACInError(boolean)

setIncludeACInError

public void setIncludeACInError(boolean newDoIncludeACInError)

Include correction to the uncertainty from block correlation function as prescribed in Kolafa, Jiri(1986) 'Autocorrelations and subseries averages in Monte Carlo Simulations', Molecular Physics (59) 1035

Parameters:

newDoIncludeACInError - auto-correlation is accounted if true, otherwise it is not.