CCD Options
Replicates of factorial points: This is the number of times each factorial run will be performed. Replicating the factorial point improves the estimates of the model coefficients, but will increase the axial distance required to achieve the rotatable and orthogonal properties listed below.
Replicates of axial (star) points: This is the number of times each axial run will be performed. Replicating the axial points improves the estimates of the model coefficients, and will reduce the axial distance required to achieve the rotatable and orthogonal properties listed below.
Center points: This is the number of center points that will be in the design. The default is the minimum suggested number of center points. Reducing this number can severely impact the prediction precision of the interior of the experiment. Increasing the number of center points improves the prediction precision, while increasing the axial distance required to achieve the orthogonal quadratic property.
Alpha values: The distance from the center to place the axial runs in coded scale.
Rotatable (k<6): the default setting for up to 5 factors, this creates a design that has the standard error of predictions equal at points equidistant from the center of the design.
Orthogonal blocks: Only present when the CCD is built with blocks. The use of this axial setting causes any block effects to be estimated independently of model coefficients.
Spherical: This puts all factorial and axial points on the surface of a sphere of radius = square root of k (the number of factors)
Orthogonal Quadratic: The alpha distance where the quadratic terms are independently estimated from the other terms.
Practical (k>5): This is the default for designs that have 6 or more factors. The alpha value is the 4th root of the number of factors. This has been shown to produce axial values that can practically be run, and yet the design still has sound statistical properties.
Face Centered: Pull the axial points into the faces of the cube - at +/- 1 levels. This produces a design where each factor only has 3 levels. It is used when the area of interest is nearly as large as the area of operability.
Other: Specify any alpha value desired. We recommend you check Design Evaluation before completing the runs.
Suggestion: If the practical alpha is too large consider using “Other” with a value of no less than 1.5. This alpha distance will provide a more realistic range for the experiment than practical alpha while providing better estimating properties than a face centered alpha range.