Mixture Designs for use in STRATEGY for Windows

User says: Lots of my problems involve mixtures, either alone or in combination with process factors. I could generate designs using STRATEGY for DOS, but it is not apparent how to do this in STRATEGY for Windows.

Dave Doehlert currently recommends three types of mixture designs. Other types, developed in the 1970's, are no longer competitive. (Note that STRATEGY for Windows can analyze any quality design.)

1. Scheffe Simplex Designs. These designs apply when there are no process variables involved and each component can vary from zero to one hundred percent of the mix. These designs will be included in an upgrade of STRATEGY for Windows sometime in the first quarter of 1997.
   First, note that if only two components (A and B) are involved in your experiment, then percent of A is a factor in the ordinary process variable sense. Use a standard process factor design with percent of A as a factor; B does not need to be included in the design. The level of B is then implied as one hundred percent minus the percent of A.
   Note that each "component" (A, B, C, etc.) can itself be a mix of certain amounts of other components when the commercial interest is only in mixtures of A, B, C, etc. For this reason, Scheffe Simplex designs also apply to mixtures with lower limits only.
2. When both an upper and lower limit can be set on each component and the sum of the upper limits is less than or equal to 1.0, use all of the components but one as process factors in an ordinary process factor design. Add one more column for that last mixture component, and in that column place 100% - the sum of the other mixture components.
   It is not important which component is listed last. When defining the model file, the last component must be left out of the model for computation to proceed.
3. If your mixture problem is not one of the types above, you need a custom design. Fill out the on-line form or fax your components, limits, and constraints to The Experiment Strategies Foundation at 800-775-3751. They will respond with a price for a design which maximizes precision while holding the number of experiments down to within your budget limit.
   Your project will probably involve both process and mixture variables, and could very easily involve discrete (qualitative) factors also. All of these should be studied in one design to benefit from the interactions. The Experiment Strategies Foundation will compute a design for you which studies all of these types of factors.

STRATEGY for Windows Knowledge Base Index

STRATEGY for Windows Home Page