Maria Lanzerath, Statistician, WL Gore & Associates
In multifactor designed experiments, there are sometimes factors involved that have a high impact on the outcome and cannot be fully randomised. As a result, blocking of one or more process factors is required. This changes the structure of the DOE, in planning as well as in the analysis. If this different structure remains ignored, the amount of gathered information is reduced, and powers and significances are simply wrong and may lead to false conclusions. The goal of this talk is to create awareness on when a split-plot DOE is appropriate, and to show how the planning and analysis are done. Several examples using JMP will be shown.