9624469 Nakayama This project involves the development of asymptotically valid simulation procedures for comparing the steady-state behavior of a small number (less than 20) of different systems. Most steady-state comparison results for simulation assume that the underlying populations are normally distributed, and that observations from each population are independent and identically distributed. These assumptions are violated in many practical situations, particularly in complex dynamic systems, where observations may be highly correlated and not normally distributed. This research develops simultaneous confidence intervals for steady-state parameters of stochastic processes under a much weaker assumption, namely that the underlying stochastic process satisfies a functional central limit theorem. The research focuses specifically on developing simultaneous confidence intervals for pairwise comparisons of means, contrasts, linear combinations of means, multiple comparisons with a control, and multiple comparisons with the best. A software package will be developed that incorporates the new techniques developed. The educational component entails the development of two new courses and simulation and modeling laboratory. The development of efficient techniques for comparing simulated systems can have a great impact on a wide spectrum of industrial problems, including many large-scale problems in manufacturing and telecommunications where simulation may be the method of choice for analysis and decision-making.
