Resumen de A Split-Plot Experiment with Factor-Dependent Whole-Plot Sizes

Eric Schoen, Bradley Jones, Peter Roos

• The FrieslandCampina dairy company wished to redesign its coffee cream production process while maintaining the same viscosity for the product. Investigators wanted a model linking nine controllable factors, two of which were hard to change (HTC) and seven of which were easy to change (ETC). Because of the combination of HTC and ETC factors a split-plot experiment was necessary. The number of runs within a plot depended on the level of one of the whole-plot factors, a situation not covered by commercially available software. This paper details three existing algorithms that can be modified to produce optimal split-plot designs with factor-dependent whole-plot sizes. One of these was used to design the FrieslandCampina experiment to study the effect of process design changes on coffee cream viscosity.