ikr.simlib.statistics.boundary

Class BoundaryStatistic

java.lang.Object

ikr.simlib.statistics.Statistic

ikr.simlib.statistics.boundary.BoundaryStatistic

All Implemented Interfaces:

Printable

Direct Known Subclasses:

JainBoundaryStatistic, StdBoundaryStatistic, UnlimitedUnbatchedBoundaryStatistic

public abstract class BoundaryStatistic
extends Statistic
implements Printable

BoundaryStatistic is an abstract base class derived from Statistic. It is the base class of StdBoundaryStatistic and JainBoundaryStatistic. These two classes are used to find the quantile for a given percentage. The class StdBoundaryStatistic provides an implementation to calculate quantiles for a given number of samples and percentage. The class JainBoundaryStatistic does the same with an alternative method, the P2 algorithm described in "R. JAIN, I. CHLAMTAC: "The P**2 algorithm for dynamic calculation of quantiles and histograms without storing observations." Communications of the ACM, Vol. 28, Oct. 1985, pp. 1076-1085.". Thus, for constructor call, StdBoundaryStatistic requires additionally to the percentage (e.g., 0.95) also the expected number of samples.

The following functions are available for evaluation:

• getQuantile() returns the value of the quantile.
• getPercentage() returns the value of the input percentage.
• getNoOfSamples() returns the number of all collected samples.

Boundary statistics can't be used with batch parallelization.

Following keywords are defined for printing the results:

quantile

value found to be the quantile

prob

input percentage

numberofevents

number of events in all batches

A special case of a quantile is the median, which is the 50%-quantile. The dedicated MedianStatistic has been implemented to measure medians.


Accuracy of the median algorithm
The graph in Figure 1 shows the accuracy of the median approximation in comparison to the accurate value when define the statistic with negative exponentially distributed values. A total of 10,000 samples have been used in this test.
Tests have shown that a MaxBufferSize of 10-20% results a deviation to the accurate value about 0.01%.

Field Summary

Fields inherited from class ikr.simlib.statistics.Statistic

sampleIndex, simNode, traceWriter

Constructor Summary

Constructors

Constructor and Description
BoundaryStatistic(SimNode ownNode)

Method Summary

Methods

Modifier and Type	Method and Description
abstract int	getNoOfSamples()
abstract double	getPercentage()
abstract double	getQuantile()
java.lang.String	getResultTypeName()
void	printResults(ResultType type, PrintResultWriter writer)
abstract void	update(double sample)

Methods inherited from class ikr.simlib.statistics.Statistic

addResultTag, addResultTag, computeMeasures, disableTracing, enableTracing, handleInitSimulation, handleStartBatch, handleStartTransientPhase, handleStopBatch, handleStopTransientPhase, printComments, resetBatchStatistic, resetStatistic, update, writeTraceEntry

Methods inherited from class java.lang.Object

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

Constructor Detail

BoundaryStatistic

public BoundaryStatistic(SimNode ownNode)

Method Detail

printResults

public void printResults(ResultType type,
                PrintResultWriter writer)

Specified by:

printResults in interface Printable

getResultTypeName

public java.lang.String getResultTypeName()

Specified by:

getResultTypeName in interface Printable

update

public abstract void update(double sample)

getQuantile

public abstract double getQuantile()

getPercentage

public abstract double getPercentage()

getNoOfSamples

public abstract int getNoOfSamples()