Split-Plot Central Composite

Split-plot central composite designs (CCD) are different than randomized CCD’s because some of the factors are set as hard to change (HTC). HTC factors are too difficult or expensive to reset every run. When the design is built the runs where the levels of the HTC are the same are grouped together. The end result is fewer changes are necessary for the HTC factors. The downside when restricting the randomization is that more runs are required and the precision of the model is worse than if the design had been randomized.

Numeric Factors: How many continuous numeric factors are involved in the experiment?

For split-plot CCD’s categoric factors and blocks are not available. Use split-plot optimal designs when you have categoric factors and/or blocks along with the numeric factors.

Enter factor ranges in terms of …
  • +/-1 to define the limits for the area of interest where the optimum is believed to exist. Axial points will typically be outside this limit. Enter the limits in the Low and High columns.

  • alphas to ensure that even the extreme axial runs are within the area of operability. The area of interest must be within the area of operability. Enter the limits in the -alpha and +alpha columns.

Type: There will be different options available in this pull down based on how many factors are to be included in the experiment. The designs listed here form the factorial core of the central composite design. The default suggestion is the largest fraction that will produce a design under 1000 runs or maintain at least resolution V behavior.

Options: Click on the options button to change the axial (alpha) distance which is how far the star points will be from the center in coded units. Use the help button on the options dialog for more information.

The Points area provides a preview of the number and type of runs that will be in the design.


See central composite designs for a more general description of this type of experiment.