Detecting shifts in the mean of a simulation output process

The application of the correct simulation output analysis technique requires a knowledge of the model's behaviour. Traditionally, only two types of model behaviour are discussed for discrete event simulations: transient and steady state. In this paper, a third type of behaviour, denoted shifting steady state, is considered in which the model passes through successive periods of different steady states. A heuristic technique for identifying the shifts in the mean of a time series is applied to the output data from simulation models with known shifting steady-state behaviour in order to test its effectiveness in detecting the shifts. The heuristic performs well, indicating that it may be a valuable additional technique for output analysis.