I have the following question:
I am working with a fractional factorial design 2^(5−1), evaluating a single response, and adding a central point and a replicate to the effects. The goal is to determine if there are indications of curvature and if there is a difference between replicates, respectively.
In the effect summary table, if I exclude the replicate effect and exclude curvature, the model does not show a lack of fit. If I remove curvature effect, the values does not changes, but if I remove replicate effect, the model show a lack of fit. What is the difference?
Is there a correct approach? Should I exclude or remove non-significant effects?