More on Mixture Designs

Creating a Constrained Mixture Design

ADX eanbles you to place linear constraints on the factors in addition to the basic constraint that the factors sum to one. Each linear constraint can involve one or more factors.

Since there is no standard design covering all situations for constrained mixtures, ADX uses the optimal design interface for the creation of these designs. This interface chooses runs from the vertices of the constrained region, the centers of the edges, the centroids of the faces, and the overall centroid. You can choose a design for a linear, quadratic, special cubic, full cubic, or quartic model, depending on how many factors are present in the model.

Fitting a model to a constrained mixture design involves the same steps as fitting a model to an unconstrained mixture design. Furthermore, the optimization tools available in analyzing an unconstrained mixture design are also available for the constrained design. The only difference is that ADX marks the feasible region defined by the linear constraints and prevents factor settings outside the bounds of those constraints.

The following sections show how to create a mixture design with three factors, X1, X2, and X3, with the following constraints:

• X1 0.1
• X2 2 X3

A quadratic model will be used.

---

Defining Variables and Constraints

Creating the Design

Analyzing and Optimizing a Constrained Mixture Experiment