Definitive Screening Design

Definitive Screening Design (DSD) is a three-level screening design that can simultaneously estimate main effects, quadratic effects, and some two-factor interactions — introduced by Jones and Nachtsheim (2011), DSDs overcome the limitations of traditional 2-level screening designs.

DSD Advantages

• Three Levels: Includes center points for each factor, allowing detection of curvature.
• Structure: $2k+1$ runs for $k$ factors (e.g., 13 runs for 6 factors).
• Model Building: With ≥6 factors, DSDs can estimate a full quadratic model in active factors.
• No Confounding: Main effects are completely unconfounded with two-factor interactions.

Why It Matters

• Screen + Model: Combines screening and response surface modeling in a single design — saves an experimental round.
• Curvature Detection: Unlike Plackett-Burman, DSDs detect non-linear effects that indicate you are near an optimum.
• Modern Standard: Rapidly becoming the preferred screening design in practice (available in JMP, Minitab).

DSD is screening and optimization in one shot — a modern design that identifies important factors AND models their effects simultaneously.