Simple Multivariate Time Series for Simulations of Complex Systems

Abstract

Recent work has made the generation of univariate time series for inputs to stochastic systems quite simple. The time series are all random linear combina- tion~ of i.i.d. random variables with Exponential, Gamma and hyperexponential marginal distributions. The simplicity of structure of these time series models makes it practical to combine them to model multivariate situations. Thus one can model, for example, alternating sequences of response and think times at a terminal in which response and· think times are not only autocorrelated, but also crosscorrelated.