Introduction to Space-Filling Designs Mini-Paper from the October 2013 Newsletter
Abstract: Design of Experiments (DOE) is a powerful tool for measuring and improving processes and designs that allows researchers to maximize the amount of information collected from an experiment while minimizing the cost of the experiment. Students of Statistics and Quality disciplines are usually acquainted with several different types of DOEs, e.g., factorial, response surface, and split-plot designs. Experimental designs of these types were developed with real-world experiments in mind, where determining which factors drive an output is more important than making an accurate predictive model of the phenomenon. As computers become increasingly powerful tools for modeling phenomenon which cannot be produced in a lab, new designs of experiments need to focus more on accurately capturing a modeled response, regardless of the complexity of the issue. As a result, a new class of DOE, called Space-Filling Designs, is gaining popularity in a variety of fields including engineering and environmental science. This paper explains the motivation for using a Space-Filling Design as well as the process for creating a Space-Filling Design. Also, the pros and cons of Space-Filling Designs versus those of traditional DOE types are summarized.
Keywords: Designed experiments - Space-filling designs - Latin Hypercube Designs - LHD - Computer Experiments