Modelling input distributions in discrete event simulations

Discrete-event simulation almost invariably makes use of random quantities drawn from given probability distributions to model chance fluctuations. Such distributions are called input models and this tutorial discusses how to choose appropriate input models. Factors which should be considered in selecting input models are discussed. Input models depend on parameters that often have to be estimated from real data. We show how the parametric bootstrap can be used to investigate the distributional properties of parameter estimators and also to give a goodness-of-fit test thereby providing a method for selecting between competing candidate input models.