On the use of variance reduction techniques when comparing simulation systems at the steady state

The analysis of non-terminating simulations requires accurate point and interval estimation for measures of performance at the steady state. Variance reduction techniques are intended to improve precision in the estimates. Many of these methods have been focussed on the study of a single system. However, little attention has been devoted to the problem of applying variance reduction techniques when comparing different systemsʹ configurations at the steady state. A first step in addressing this problem is determining the feasibility and convenience of using a specific variance reduction technique. In this work, an analysis and recommendations for using several well-known variance reduction techniques on different situations, including cases where the systems do not reach the steady state, are presented.