Experimental optimization of statistical simulation

Although simulation involves solving a mathematical model through experimentation, the literature of simulation does not reflect a broad, sustained interest in the design of simulation experiments. The objective of this paper is to fulfill a part of this need by suggesting a general five-phase experimental approach for determining the optimum for a particular, but common, type of simulation problem. The majority of the suggested procedure makes use of existing design and analysis techniques. However, the problem of multimodal simulation outcomes has resulted in the development of the “random factorial” experimental design. This design is a combination of a complete factorial design and a random balance design. The majority of the paper is devoted to a discussion of the use of this design approach for solving the multimodal problem.