A0271
Title: Criteria for assessing space filling of a design with emphasis on the stratification pattern
Authors: Ulrike Groemping - Berliner Hochschule fuer Technik (Germany) [presenting]
Abstract: Designs for computer experiments in quantitative factors should fill the experimental space as well as possible. There are many criteria for the assessment of space-filling properties of a design, and the stratification pattern is a recent addition to this toolbox. It is based on coding for qualitative factors, which renders its calculation computationally demanding. The pattern can be broken down into dimensional contributions, which provide an additional understanding of the design's properties. The logic behind the stratification pattern and its breakdown into dimensions by weight tables are presented. A small simulation study compares the prediction performance of several designs for which the stratification pattern and several further space-filling criteria are considered. Its partly surprising outcomes indicate the need for further research regarding the choice of a design (resp. the choice of a criterion for choosing a design) for actual experimentation.
