Custom designs can combine up to two sets of mixture components (e.g. cake and frosting) with numeric (e.g. time in and temperature of the oven) and categoric (e.g. round or square shape) process variables. They can also focus on a single type of factor. The are called “custom” due to their flexibility.
All of the custom designs can accommodate additional constraints on the components and factors.
Entire mixtures and individual factors can be set as Hard to change to create a split-plot design.
Optimal designs are built algorithmically with a purpose to fit a specified design model (quadratic by default). The algorithm can freely search for the best points or be restricted to a candidate set.
User-Defined designs include every run from a candidate set.
A candidate set can come from the internal candidate list, discrete factor settings, or from a predefined candidate file (*.cndx).
Historical designs are used to create a design framework for data that already exists. The result is an empty design layout spreadsheet where the existing data can be typed or pasted.
For all three of the above, evaluation, analysis and optimization proceed as with any other design.
Simple Sample designs have no factors. They are a way to compute descriptive statistics, check for normality and plot the observations against the mean and interval estimates.