The role of statistical methodology in simulation

Abstract

In practice, a most popular method to study complex systems is simulation. However, a major practical drawback of simulation is its ad hoc character. After spending much mental effort and computer time to develop, program, and run a computer simulation, strictly speaking, the results are valid only for the specific parameter values and structural relationships of the executed simulation program (or programs). Changing a parameter or relationship means that the simulation program has to be run again. Yet, such changes are necessary to answer "what if" questions, to find optimal system configurations and, not to be forgotten, to establish the sensitivity of the conclusions to specific model assumptions.