What is a Definitive Screening Design?
A Definitive Screening Design (DSD) allows you to study the effects of a large number of factors* in a relatively small experiment. In simple terms, DSDs are an improvement on standard screening designs (like the Plackett-Burman) that prevent confounding of factors and can also detect non-linear responses.
Definitive Screening Design vs. Standard Screening Design
One of the main differences (SAS, 2015) is that a definitive screening design can estimate quadratic (curvilinear) effects when the model contains only main effects and quadratic effects; a standard screening design can only detect linear effects on the response.
Another significant difference between the two designs is how they handle confounding. In Definitive Screening Designs, there is never complete confounding with any sets of two-factor interactions. For standard screening designs, the issue of confounding is only addressed with fractional factorial designs that are Resolution IV or higher. Resolution tells you how much confounding is in your model; resolution IV designs have—at most—confounding involving three-factor interactions. Confounding in resolution III designs (which are probably used the most) isn’t addressed at all with a standard screening design.
Appropriate Uses for DSDs
- Continuous factors are ideal.
- Categorical factors are strongly not preferred. A small percentage of categorical factors are fine, but too many and the result is what Jones and Nachtsheim (2013) call “…an undesirable choice.”
- Start with at least six factors . For 4 or 5 factors, create a DSD for 6 factors, then drop the last two columns.
- Each individual factor should be able to combine with any other factor. If any combinations are not possible, this reduces the design’s power.
- Factors should be independent of each other. For example, if one goes up, it shouldn’t cause another factor to go down.
- Use during the first experimental stages when reducing a large number of factors to a more manageable number.
- Do not run the DSD as a split-plot design or when the a priori model has higher order effects.
Jones, B. & Nachtsheim, C. (2013). Definitive Screening Designs with Added Two-Level Categorical Factors. Journal
Journal of Quality Technology. Volume 45, Issue 2. Retrieved December 10 2017 from: https://www.jmp.com/content/dam/jmp/documents/en/white-papers/dsd-with-added-two-level-categorical-factors.pdf
Jones, B. (2016). Proper and improper use of Definitive Screening Designs (DSDs). Retrieved December 10 2017 from: https://community.jmp.com/t5/JMP-Blog/Proper-and-improper-use-of-Definitive-Screening-Designs-DSDs/ba-p/30703
SAS (2015). JMP 12 Design of Experiments Guide. SAS Institute.
If you prefer an online interactive environment to learn R and statistics, this free R Tutorial by Datacamp is a great way to get started. If you're are somewhat comfortable with R and are interested in going deeper into Statistics, try this Statistics with R track.Comments? Need to post a correction? Please post on our Facebook page.