Package de.uka.ipd.sdq.statistics

Class PhiMixingBatchAlgorithm

java.lang.Object
de.uka.ipd.sdq.statistics.ABatchAlgorithm
de.uka.ipd.sdq.statistics.PhiMixingBatchAlgorithm
All Implemented Interfaces:
IBatchAlgorithm
public class PhiMixingBatchAlgorithm extends ABatchAlgorithm
Implements a batch means procedure based on phi-mixing conditions as described in [1]. Appropriate batch sizes and the number of batches are determined automatically.

The procedure utilizes an independence test in order to build a so-called "quasi independent" (QI) sample sequence. By default the RunUpTest will be used. "The aim of the QI method is to continue the simulation run until we have obtained a pre-specified number of essentially independent random samples by skipping highly correlated observations." [1] As soon as the QI sequence appears to be independent, the computed batches can be considered as valid. Samples in the QI sequence are only used to determine appropriate batch sizes. They are not used to compute the batch means! Instead, the batch means consist of all samples, regardless of statistical dependence.

[1] E. Chen, W. Kelton: A Stopping Procedure based on Phi-Mixing Conditions. Proceedings of the 2000 Winter Simulation Conference.

  • Constructor Details

    • PhiMixingBatchAlgorithm

      public PhiMixingBatchAlgorithm()

    • PhiMixingBatchAlgorithm

      public PhiMixingBatchAlgorithm(IIndependenceTest test)

  • Method Details

    • offerSample

      public void offerSample(double value)