Evaluation Model Screen

Before you evaluate a design, you must decide what model to evaluate. This model will be different, depending on what information you want to look at in the evaluation report. The list of model terms shows which terms are in the model model, which are in the error error, and which are excluded excluded.

Here are some examples:

  1. Evaluate aliasing and statistics other than power for a fractional factorial design before any runs have been made – choose the order to be 2FI (two-factor interaction). By default, the software puts four-factor interactions and higher to the ignore status. This means that when the alias structure is displayed, only main effects, 2FI’s, and 3FI’s will show. Use the Design Model selection if you have built a special design with unusual terms.

  2. Evaluate power for a fractional factorial – change the order to Main Effects. The power calculation needs to be done on the expected number of significant effects. Although we don’t know this information ahead of time, a reasonable guess is that the number of significant effects will be approximately equal to the number of factors.

  3. Evaluate a design after the experiment has been run – the software will default to the model that the design was originally set up for (i.e. a CCD would default to a quadratic model). You also have the option of evaluating a specific response. This is handy when some of the responses have missing data.

  4. Evaluate a response surface or mixture design – choose the model order to reflect what the experiment was designed for – a quadratic, special cubic, etc.