Chapter 13 - STATISTICAL AND ANALYTICAL MODELS IN SIMULATION

Abstract

Most computer simulations are statistical sampling experiments performed on a model of the system under investigation. Among the issues involved in the statistical design and analysis of simulation experiments are initial conditions, run length, the statistical reliability of simulation estimated parameters, variance reduction techniques, the generation of random variates, and experimental layout. The simulation models of regenerative systems have the property of starting afresh probabilistically from time to time; thus, a simulation run can be divided into independent and identically distributed cycles. This results in a simple solution to the initial conditions problem and run length determination problem. Another approach for achieving higher precision within a fixed length of a simulation run is to reduce the variances of the important estimators in the simulation at the expense of increasing the variances of less important estimators.