Multivariate standardized time series for steady-state simulation output analysis
成果类型：
Article
署名作者：
Muñoz, DF; Glynn, PW
署名单位：
Instituto Tecnologico Autonomo de Mexico; Stanford University
刊物名称：
OPERATIONS RESEARCH
ISSN/ISSBN：
0030-364X
DOI：
10.1287/opre.49.3.413.11209
发表日期：
2001
页码：
413-422
关键词：
摘要：
The theory of standardized time series. initially proposed to estimate a single steady-state mean from the output of a simulation, is extended to the case where: more than one steady-state mean is to be estimated simultaneously. Under mild assumptions on the stochastic process representing the output of the simulation, namely a functional central limit theorem. we obtain asymptotically valid confidence regions for a (multivariate) steady-state mean based on multivariate standardized lime series. We provide examples of multivariate standardized time series, including the multivariate versions of the batch means method and Schruben's standardized sum process. Large sample properties of confidence regions obtained from multivariate standardized time series are discussed. We show that, as in the univariate case. the asymptotic expected volume of confidence regions produced by standardized time series procedures is larger than that obtained from a consistent estimation procedure. We present and discuss experimental results that illustrate our theory.