Candidate points are used to limit the factor and component combinations available to the experiment. Optimal and User-Defined designs have an option to use a candidate set when building the design.
There are three types of candidate sets available in the Stat-Ease ® software: discrete factor type, external candidate files, and the internal candidate list.
Discrete Factor Type
When specifying the factor information, The type column of numeric (non-mixture) factors can be set to Discrete. This will activate the levels column. Type in the number of levels and then the settings to be used for that factor. Building designs with discrete factors does not actually create a candidate file, but does limit the software to only choosing factor settings specified by the discrete settings.
Shortcut Trick: if the levels will be evenly spaced, set L[1] to the lowest value and L[2] to the highest value before entering the number of levels.
External Candidate File
External candidates sets come from a list of viable factor combinations provided by the experimenter.
External candidates must include all the factors. It is not possible to specify a candidate for a subset of the factors and have the remainder be discrete, continuous, or come from the internal candidate list.
Most often, a Historical design is created to hold the list; although, any design can be used to get things started. Once the design is populated, it is saved as a candidate file by clicking on the Design Tools menu and selecting Write Candidate File. Chose the directory and file name to save the *.cndx file.
External candidate files (.cndx) must be read in to the design.
To use an external candidate file in a user-defined design, click the Read list button and navigate to the where the file was saved.
For an optimal design the Read list button is under the Edit candidate points button.
Internal Candidate List
The internal candidate set includes vertices, centers of edges, centroids and other common space types. These settings allow very good estimates for a quadratic model and reasonable estimates for a full cubic model.
Vertices are at the extreme limits of the design.
Center of edges are half-way between two vertices.
Thirds of edges are one-third and two-thirds the way along an edge between two vertices.
Triple blends are a third along two adjacent edges putting them on a constraint plane.
Constraint plane centroids are the center of a hyper-plane defined by three or more coplanar vertices.
Axial check points/blends are half-way between the overall centroid and the vertices.
Interior points/check blends are half-way between the centroid and center of edges.
Overall centroid is the geometric center-of-mass of the design.