Chapter 17 - A VARIANCE REDUCTION TECHNIQUE THAT INCREASES REGENERATION FREQUENCY

Abstract

This chapter discusses a class of methods designed to increase the computational efficiency of computer simulations. This increased efficiency is obtained by simulating a stochastic process that is related to, but different from, the stochastic process of interest. Each event in the new process will correspond to several events in the original process. The simulation of this new process will then, in some sense, proceed at a faster rate than that of the original process. One of the main difficulties with regenerative simulations is that even though it may be known that the process being simulated is regenerative, the regenerations may be few and far between. Thus, even for very long simulation runs, only a relatively few identically distributed (i.i.d.) cycles are observed. One possible remedy to this problem is the use of approximate regenerations. The idea is to break the simulation run up into blocks that are almost i.i.d. by defining a sequence of almost regenerations. These blocks are then treated as if they are i.i.d.