Simulation analysis with group screening

Abstract

Listen

Translate article

Simulation of systems involve a number of factors or variables which act and interact to produce an output. When these factors are small in number the experimenter can control them effectively to obtain an optimal output, but when the number of factors increases then efforts to control all the factors effectively also increases and leads to wastage of resources and away from optimality. Very often in a system where a number of factors are involved only a few of them actively involve in producing the output. Therefore it is the interest of the researcher to detect those active factors, and experimental design is frequently applied. The detecting process is called as screening. Screening can done by factorial designs such as 2 K (full factorial design) or by 2 k-p (fractional factorial design). Even by the fractional factorial method, one might still need to run too many experiments. It is desired to accomplish the screening process in as few number of runs as possible using random balance design, super saturated design or group-screening. Of these methods group-screening has been identified by researchers such as Mauro, Smith, etc. as the most efficient tool. In this research we study the usage of group-screening method as a simulation analysis tool while improving the method from earlier researches. Performance and results are studied and provided in form of response surface figures and tables.