Regenerative simulation of TES processes (Q1323530)

The authors focus on a versatile class of Harris-recurrent Markov chains, called TES (Transform-Expand-Sample). TES processes can generate a variety of sample paths with arbitrary marginal distributions, and autocorrelation functions with a variety of functional forms (monotone, oscillating and alternating). A practical advantage of TES processes is that they can simultaneously capture the first and second order statistics of empirical sample paths. TES modeling methodology can simultaneously match the empirical marginal distribution (histogram), as well as approximate the empirical autocorrelation function. The autocorrelation function of a TES process can be computed from fast numerical formulas without requiring simulation. Furthermore, Monte Carlo generation methods of TES sample paths are very fast and have longer periods than the underlying pseudorandom number generator. These favorable computational properties are important in interactive modeling, since the numerical computations are much faster than the corresponding calculations of simulation-based statistics. A brief summary of the regenerative method and some of its ramifications is proposed. The special case of Harris chains is surveyed with special emphasis on the identification of regeneration points in special systems such as queues. A construction procedure of TES regeneration points is defined. The paper contains a special case study of a G/GI/1 queue, which is denoted by TES/M/1, where the likelihood ratio method is used to calculate the mean waiting time performance as a function of the intrinsic TES parameter controlling burstiness (magnitude of autocorrelation) in the arrival process, and the score function method is used to estimate the corresponding sensitivity with respect to the service rate.