ABSTRACT
Simulation models rely on specification of probability distributions to characterize the variability or uncertainty in key parameters. This chapter explains a variety of ways of choosing the appropriate distributions. The first type of distributions is that used for modeling expert opinion. It is simple conceptually, but we provide guidance on best practices for soliciting expert judgments, and we show how to aggregate multiple expert opinions. The second method is fitting distributions to data. When relevant data is available, it is desirable to use it, and we provide examples illustrating visual and quantitative measures of fitted distributions. The final method for selecting distributions is based on the idea that the math of different distributions reflects real-world situations, for example, a Binomial distribution that represent tossing a coin multiple times. Other examples are the Poisson, Exponential, Weibull, Lognormal, and Gamma distributions that can be appropriate choices for particular types of uncertain variables. We provide several examples and general guidance for these choices. Chapter examples include a company's valuation, a value at risk (VaR) analysis, and a cockpit component failure simulation.
