Dealing with Aliased Models

If you select a model with more terms than the number of unique points in the design, some terms will be aliased. This can happen intentionally (for example, most RSM designs are too small to estimate cubic models) or unintentionally by not completing all the experimental runs (missing data). To keep you from being misled, Stat-Ease flags aliased models with warning messages in several places.

An aliased model should not be used for the prediction of a response. However, by reducing the number of terms in the model (eliminating insignificant terms) you may be able to find an adequate model without aliased terms. In Stat-Ease you can edit the model to remove terms that may not be important. This is done by toggling the individual terms on the Model or Effect List screen to either keep them in the model or exclude them.