ABSTRACT
This chapter explores how to model relationships between variables, both inputs and decisions, in simulation models. Some relationships can be modeled using logical statements and functions (IF, AND, OR, etc.). Others utilize a variety of correlation structures. We provide examples of a several techniques and methods, including nonlinear correlations (a variety of copulas) and the use of correlation matrices to accommodate many interrelated model components. We will also show how to use historical data to fit an appropriate correlation structure, as well as how to implement assumed correlation patters and investigate the sensitivity of model results to these assumptions.
The chapter also includes a section on regression analysis. Regression models can also capture how several variables jointly affect another factor in a simulation model. Inherent in regression modeling is that the model fit has uncertainty, and this uncertainty can be quantified and used in simulation modeling. The examples in this chapter include drug development, collateralized debt obligations, system component failures, and advertising effectiveness.
