Optimization Options

The defaults provide good results and should generally be used.

Random Starting Points: This controls the number of random starting points used to start the simplex optimization search. These points can come from any where in the design space. Increasing this number will slightly improve the likelihood of finding the global optimum.

Use up to X Design Points: In addition to the random starting points the software will include up to X of the design points. These design points are selected at random from the list of points actually tested. 50 is enough points to handle most designs.

Duplicate solution filter: This controls the difference between response values that is considered to be “the same” solution. Moving the slider bar to the right will make a larger difference “the same”, thus reducing the number of nearly identical solutions. Moving the slider bar left will reduce the difference, causing an increase in the number of nearly identical solutions.

Simplex fraction: The simplex optimization works by stepping across the factor space looking for locally optimal solutions. Decreasing this value will make the simplex pattern smaller, covering the design space in smaller increments and taking more time to reach an optimum. Increasing the value will make the simplex have a larger pattern, covering the space more quickly, but at the risk of over-stepping a local optimum.

Maximum number of solutions: The number of solutions can grow very large. This limits the number shown.

Include Standard Error Models: Turning on this box will add a StdError response for each response in the design. It can have goals set just like any other response. It is the standard error for the average prediction at the factor settings being considered by the optimizer. A common use of this option is to limit how far factor ranges can be extrapolated without generating less than useful predictions.