Qualitative factor with all rows for a certain setting excluded excludes that term (Q916)
In the listed versions, MODDE handles condition number and fitting a model for a design with rows/experiments for a qualitative setting excluded.
Qualitative factors in MODDE are expanded into K-1 terms in the calculations. For a qualitative factor with f.i. 4 settings, 3 terms are created for calculations. When you exclude all rows for one of the settings, the corresponding term becomes constant and the matrix ill conditioned. From MODDE 11, this constant term is automatically excluded and calculations performed as if that qualitative setting was never in the design.
This is done automatically without notification, except for the case where the model is already fitted with MLR when the rows are excluded. In that case, a message about ill conditioned matrix is displayed and a question about whether to instead fit with PLS.
In this case, the procedure you need to follow is:
- Click No to not fit with PLS.
- Click OK on the info-message that the model is not fitted.
- Click Fit model to fit the model (with MLR).
- Click Yes to refit the model. The refit is needed to make sure that models for all y-variables are fitted.