Output Analysis Capabilities of Simulation Software

Although most simulation practitioners are aware that output analysis is of extreme importance in a discrete-event system simulation, many are not versed in the process for conducting that analysis. Hence, it is not unusual that a simulation practitioner will conduct just one replication of the simulation, and provide the resulting measure-of-performance as the estimated response. Point estimates like this can be very dangerous, since they represent but one of many possible values of an experiment when random variation is present. It is much more appropriate to provide a statistical confidence interval, as a minimal form of output analysis. This article uses an example that describes a context in which variation in output performance measures can occur. The actual output from a simulation of that example system is used to develop confidence intervals. The output features of simulation software are briefly described. This is followed by a listing of these same output features and their availability on five simulation software products. The intention of the article is to draw attention to simulation output analysis and to show that support for it is readily available in simulation software. It is not the intention of the article to compare the five products, or to indicate that other simulation software products do or do not have the same or similar features.