A spectral method for confidence interval generation and run length control in simulations
Communications of the ACM - Special issue on simulation modeling and statistical computing
Simulation output analysis: a wavelet-based spectral method for steady-state simulation analysis
Proceedings of the 35th conference on Winter simulation: driving innovation
ASAP3: a batch means procedure for steady-state simulation analysis
ACM Transactions on Modeling and Computer Simulation (TOMACS)
Performance evaluation of a wavelet-based spectral method for steady-state simulation analysis
WSC '04 Proceedings of the 36th conference on Winter simulation
| Hi-index | 0.00 |
We summarize an experimental performance evaluation of WASSP and the Heidelberger-Welch (HW) algorithm, two sequential spectral procedures for steady-state simulation analysis. Both procedures approximate the log-smoothed-periodogram of the batch means after suitable data-truncation to eliminate the effects of initialization bias, finally delivering a confidence-interval estimator for the mean response that satisfies user-specified half-length and coverage-probability requirements. HW uses a Cramér-von Mises test for initialization bias based on the method of standardized time series; and then HW fits a quadratic polynomial to the batch-means log-spectrum. In contrast WASSP uses the von Neumann randomness test and the Shapiro-Wilk normality test to obtain an approximately stationary Gaussian batch-means process whose log-spectrum is approximated via wavelets. Moreover, unlike HW, WASSP estimates the final sample size required to satisfy the user's confidence-interval requirements. Regarding closeness of conformance to both confidence-interval requirements, we found that WASSP outperformed HW in the given test problems.