A Note on the Effect of Initial Conditions on a Simulation Run

Abstract

A problem that is often encountered, when using a computer simulation model, is whether it is better to make a single continuous run or to make an equivalent number of independent but shorter runs. It is sometimes suggested that when the quantity of interest is a sample mean then, provided the effects of initial conditions can be neglected, it is better to make independent runs as the variance of the sample mean is then smaller. In this note it is shown that, in general, neglecting initial conditions cannot be justified and that in this case the sample mean has better properties if a single continuous run is made.