A tutorial on how to select simulation input probability distributions

An important, but often neglected, part of any sound simulation study is that of modeling each source of system randomness by an appropriate probability distribution. We first give some examples of data sets from real-world simulation studies, which is followed by a discussion of two critical pitfalls in simulation input modeling. The two major methods for modeling a source of randomness when corresponding data are available are delineated, namely, fitting a theoretical probability distribution to the data and the use of an empirical distribution. We then give a three-activity approach for choosing the theoretical distribution that best represents a set of observed data. This is followed by a discussion of how to model a source of system randomness when no data exist.