Mixture Designs

Mixture designs are used when the factors in an experiment are components in a mixture, and when you wish to determine the best proportions to use for each component. Since the proportions of all the components in a mixture must always sum to one, mixture designs assume that all the component factors are dependent (i.e., changing one component's value in an experiment will necessarily change others).

Mixture designs can also include process variables as factors. For example, if the components in a mixture are different metals that are combined to produce steel, the experimenter can choose to consider process variables like temperature and pressure in order to determine the best way to process the mixture (in addition to estimating the best proportions of the metals).

Specific mixture design types include:

• Simplex designs, which assume that there are either no bounds or only lower bounds on the component amounts.

• Extreme vertex designs are more complicated designs that allow you to add linear constraints on the design and/or upper bounds on individual components.