Validating models in public health research
The application of systems method to the understanding of public health problems (e.g., alcohol and drug abuse, chronic disease, obesity, tobacco use, and violence) has grown considerably in the past decade. System methods are seen by many of their advocates within public health as complimenting traditional behavioral and epidemiological research methods, while others see them as a fundamentally different way of understanding and explaining public health problems. Those who see the methods as complimentary often use empirical data from studies employing traditional methods and statistical analysis to validate the output of simulation models. As in other fields of applied research in which modeling has become popular, this tendency to equate a model's correspondence to data with the model corresponding to reality is especially pronounced when the goal of the modeling is to inform public policy. The present paper discusses the problems that arise when using data from an empirical study to assess the validity of a simulation model. It illustrates these problems through an examination of a specific example from the public health literature. The example demonstrates that, rather than empirical data being superior to the model, each is better considered as simply capturing a different aspect of a real-world system. Alternative means of assessing model usefulness are also discussed.
author list (cited authors)
Elkins, A. D., & Gorman, D. M.