Optimal Model Selection Screen

Search: choose from Best, Coordinate Exchange, or Point Exchange. These search algorithms are described here.

Optimality: choose from I, D, and A optimal designs. Also available and not recommended are modified distance and “pure” distance based design criterion. These criteria are described here.

Runs - Shows the number of points and the purpose for which they are selected.

Required model points is the minimum number of runs to estimate the coefficients of the terms in the designed for model specified under the Edit model button.

Additional model points are extra runs added to the experiment to improve precision estimates or coverage of the factor space.

Lack-of-fit points provide extra points to fill the factor space. The extra information provided by these points can test the fit of the model.

Replicate points of the above points are chosen to support the optimality.

Additional center points are added as requested if there are numeric factors.

Total runs: is the number of runs that will be in the final design.

Edit model: The software will default to a certain type of model, depending on the type of optimal design being created. You may change or customize the model by clicking on the Edit Model button. Adding or subtracting specific terms in a model will alter both the number of runs and specifically which runs are chosen. For more information use the help button on the Edit model button.

Blocks: allows the design to be separated into blocks. Adding blocks will add runs to your design to provide additional degrees of freedom to estimate the block effects while allowing useful analysis for the model coefficients.

Options: The more random starts and loops that are allowed the better the design will meet the chosen optimality criterion. For more details use the help button found on the Options dialog.

Force categoric balance: Available if there are categoric factors. This option forces the design to have the same number of runs for each categoric level, to the extent possible. The balance will only be perfect if the number of runs is evenly divisible by the number of categoric levels. Forcing balance may reduce the number of replicates and can make the design less optimal.

Edit Candidate Points: Click this button if want to limit the type of points the design can consider by specifying a custom built candidate set.