Hunter Screening Design

[HelenaB]

Good morning!

I found L18 Hunter Screening Design referred in an article that used JMP (in attachment). Can anyone clarify what type of design is this? I cannot found much about it in the literature.

Thank you in advance!

[Rick_SAS]

The JMP community has a separate discussion forum for JMP questions

You might want to ask your question there..

[HelenaB]

Thank you for the suggestion! I already reposted the question in that discussion forum.

[PaigeMiller]

L18 is a design that was popularized by Genichi Taguchi. It has one variable which is at three levels and the remainder are at two levels. It is orthogonal for main effects only, but interactions are partially confounded with main effects and with other interactions.

[HelenaB]

Paige Miller,

Thanks for your help. The definition that you gave is perfectly in accordance with the characteristics of the L18 Hunter screening Design that I found in the attached article. However, I also found this reference https://www.itl.nist.gov/div898/handbook/pri/section3/pri33a.htm that presents other definition for L18 Designs - 1 factor at two levels and seven factors at 3 levels. I would like to have your opinion on it.

Thank you again,

Helena

[PaigeMiller]

That L18 is 2 x 3^(7–5) design, which is orthogonal if you have main effects only. The interactions would be non-estimable, and highly confounded with main effects. I would not use this in the presence of interactions — except perhaps as a screening design to identify "active" factors, and even then one significant interaction can destroy the validity of the results from this design, this is a risk of all highly fractionated designs.

[HelenaB]

But we can say that L18 2 x 3^(7–5) design is a L18 fractional factorial design? 

[PaigeMiller · Accepted solution]

It is a mixed-level fractional factorial design. It doesn't really matter what name it has, L18 or otherwise.

[HelenaB]

Thank you for your very helpful informations!

[HelenaB]

Still regarding this answer: 

"It is a mixed-level fractional factorial design. It doesn't really matter what name it has, L18 or otherwise.

Is a L18 or other mixed-level fractional factorial design capable of generate a quadratic model, or just a main effects one? If yes, under which circunstances?