3 levels in DOEs ?

[Suhas]

Do you use three levels in DOEs?

[SeanS]

Yes, but only after I've established that I'm truly in the neighborhood of the optimum. Once I've "surrounded" the optimum, then I may even use 4 or 5 levels.

In the world of molding since we have a reasonable idea where the edges of the envelope are, we can start with more than 2 levels. In other examples, we may not have an idea where the edges are so we need to use two levels to climb the mountain, and then once we've straddled the top, we can add a third or more levels.

[Danny R]

If you don't have the time, money & resources there shouldn't be a need to do 4 or 5 levels. What I would recommend is a One-way ANOVA, first find out if the mean of one or more populations are different or not. If not, you just wasted a few hours running a 4 or 5 level DOE and a few more measuring/analyzing. If the hypothesis infers that the means are different then at that point you select your DOE.

What has worked best in my situation is a central composit design, you will have full factorial treatments and in my experience, is a very accurate design.

[Suhas]

I keep it very simple and it has worked great for me - 99% of the times - 3 factors, 2 levels. The 1% is four factors for olefins.
I would love to hear more on this topic. (I am upgrading my s'ware and would love to incorporate additional stuff if needed in there.)
Suhas