Custom designs are used when the process requires adjustments to the experiment that cannot be accommodated by a standard design.
Some of the reasons include:
- A higher than quadratic order model is necessary to estimate the response surface
- Replicates need to spread throughout the design rather than only the center point
- Difference between the high and low of all the mixture components not the same
- Mixture and process variables in the same design
- Two independent mixtures in the same design.
- Constraints in addition to the factor and/or component limits
- Too many runs in the standard designs
- Data has already been gathered
- Combinations of the above
Optimal (custom) design’s run settings are chosen algorithmically to achieve the points above, while remaining limited to as few runs as possible. There are trade offs for this flexibility. A different design is likely to be produced even if rebuilt with the same parameters.
User-Defined designs are created using all the component and factor settings in a specified candidate list. The candidate set can come from a predefined list of point types, including, vertices, centers of edges, etc. or a user created candidate set can be read in to create these designs. Use user-defined designs to ensure ALL possible component and factor combinations from the candidate points are included in the design.
Historical data creates a blank design layout to accept component and factor settings and responses from an existing data set. Design limits must be entered first; minimum and maximum settings for the components and factors.
All of the above can be set up as fully-randomized or split-plot designs. A split-plot design is created when one of the mixtures or factors is set to Hard to change (HTC).
Simple sample analysis provides a way to estimate mean, standard deviation and interval estimates and show them on a graph.