A: Recognition and Importance of Restrictions on Randomization in Industrial Experimentation

Industrial experiments are often run in split plot mode for reasons of expediency. These experiments allow the effects of the split-plot factors to be estimated with more precision than the effects of the whole plot factors. This is unfortunate when interest lies in those whole plot factors. This paper examines a philosophical motivation and methodology for using Factor Relationship Diagrams as a proactive method for developing questions that, when answered a priori to experimentation, develop understanding of those sources of variation contributing to this lack of precision. This understanding can then lead to the selection of an appropriate experiment that mitigates the risks associated with drawing conclusions regarding the whole plot factors. Examples are given.