Fit model without interaction for disallowed combination

Phil_Kay

Hi,

 

A customer has designed an experiment (attached) that includes a 4-level categorical factor, X4, and 3 continuous factors, X1, X2 and X3.

 

The combination X4 = B and X3 > 0 is not allowed. They used disallowed combinations to ensure that this constraint was obeyed. 

 

However, they would like to estimate the interaction, X3*X4, for the other levels of X4. Is there some way to code the model that would achieve this?

 

I have tried creating dummy variables for each level of X4 (X4_A ... X4_D) but that is not a very elegant solution.

 

Thanks,

Phil

Custom Design_disallowed combination.jmp

DonMcCormack

If Generalized Regression isn't an option, not sure what other things you can do except break the interaction into 1 df contrasts, which is what you did. Alternatively, you can let Stepwise do it for you. The contrasts will be different and may be less interpretable, though.

Don