Statistics PhD Defense with Moses Obiri (Dept. of Mathematical Sciences, MSU)

11/20/2020  9:00-10:00am  WebEx Meeting

Abstract:  

A space-filling design is an experimental design for which the experimental runs are uniformly scattered in the appropriate experimental design region and which permits the fitting of many potential models as well as exploring the region for optimal design factor combinations. Many experiments in industry involve mixture components or involve mixture components with process variables.  An attribute of a mixture experiment is that the experimental factors represent proportionate amounts of the mixture components and the response depends on the relative proportion of each component. For example, in the pharmaceutical industry, tablet hardness and friability may depend on the proportion of binder, lubricant, and disintegrant used as well as process variables such as particle size and mixing time.  Finding optimal combinations of mixture ingredients and process variables to produce desirable products is made more efficient by exploring the entire combined mixture and process design space. We present a new hybrid space-filling algorithm for the combined mixture/process variable experiment which will allow the fitting of potential models and explore the entire region for optimal factor combinations.