Initial transient detection in simulations using the second-order cumulant spectrum

Procedures for detecting an initial transient in simulation output data are developed. The tests use the second-order cumulant spectrum which differs from the power spectrum in that the stationarity constraint is not required for the former. The second-order cumulant spectrum can be interpreted as the nonstationary power spectrum and is an orthogonal decomposition of the variance of a nonstationary process. The null hypothesis is that the simulation output data series is a covariance stationary process. Equivalently, all estimates of the second-order cumulant spectrum in the region which excludes the estimates of the power spectrum will have an expected value of zero. The test procedures are designed to detect initialization bias in the estimation of the mean and the variance. These procedures can be extended to detect bias in the moments of cumulants of ordern, wheren>2. Results are presented from the application of the test to simulated processes with superimposed mean and variance transients and anM/M/1 queue example.

[1]  Lee W. Schruben,et al.  Optimal Tests for Initialization Bias in Simulation Output , 1983, Oper. Res..

[2]  R. Abraham,et al.  Manifolds, Tensor Analysis, and Applications , 1983 .

[3]  W. David Kelton,et al.  Initializing for bias reduction: some analytical considerations , 1988, WSC '88.

[4]  J. Goodman.  Statistical Optics , 1985 .

[5]  R. Deal.  Simulation Modeling and Analysis (2nd Ed.) , 1994 .

[6]  M. Hinich.  Testing for Gaussianity and Linearity of a Stationary Time Series. , 1982 .

[7]  Melvin J. Hinich,et al.  A Test for Aliasing Using Bispectral Analysis , 1988 .

[8]  L. Schruben.  Initialization Effects in Computer Simulation Experiments , 1983 .

[9]  Paul Bratley,et al.  A guide to simulation , 1983 .

[10]  Andrew Harvey,et al.  The econometric analysis of time series , 1991 .

[11]  W. Whitt.  Planning queueing simulations , 1989 .

[12]  James R. Wilson,et al.  A survey of research on the simulation startup problem , 1978 .

[13]  Averill M. Law,et al.  Simulation Modeling and Analysis , 1982 .

[14]  M. Rosenblatt.  Stationary sequences and random fields , 1985 .

[15]  M. Hinich.  Detecting a transient signal by bispectral analysis , 1990 .

[16]  James R. Wilson,et al.  Evaluation of startup policies in simulation experiments , 1978 .

[17]  D. Brillinger.  An Introduction to Polyspectra , 1965 .

[18]  C. J. Ancker,et al.  Evaluation of commonly used rules for detecting “steady state” in computer simulation , 1978 .

[19]  Albert N. Shiryaev,et al.  On a Method of Calculation of Semi-Invariants , 1959 .

[20]  Lee W. Schruben,et al.  Detecting Initialization Bias in Simulation Output , 1982, Oper. Res..

[21]  M. Hinich,et al.  The application of the discrete Fourier transform in the estimation of power spectra, coherence, and bispectra of geophysical data , 1968 .

[22]  H. Scheffé,et al.  The Analysis of Variance , 1960 .

[23]  Philip Heidelberger,et al.  Simulation Run Length Control in the Presence of an Initial Transient , 1983, Oper. Res..

[24]  W. David Kelton,et al.  Random Initialization Methods in Simulation , 1989 .