Interpreting Power in Mixture DOE – Simplified Taken from Fall 2006 Newsletter
Abstract: A planned mixture experiment may have very low power. The cause may be due to constraints among the mixture elements in order to have feasible and reasonable mixtures. Only interaction terms not associated with the constrained element(s) will produce reasonable collinearity coefficients and power. The paper examines the effect of replications of the mixture design to solve this problem. The result is that only the interaction term not associated with the constrained factor will have improved power, but everything else will not be improved. To examine this situation, the paper examines the 3D surfaces and the “Scheffé polynomial” associated with the design. As a final remark, the paper states that despite these limitations, models of response from the mixtures may still prove to be useful, especially as the number of blends in the experiment increase.
Keywords: Mini paper - Mixture experiment - Power - Schaffé polynomial - 3D surface - Standard error (SE) - Replicates