Using nonregular designs to generate space-filling designs

• Published in: Journal of statistical planning and inference Vol. 221; pp. 201 - 211
• Main Authors: Chen, Guanzhou, Tang, Boxin
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2022
• Subjects: Addelman–Kempthorne construction; Computer experiment; Non-empty-cell design; Strong orthogonal array; Computer experiment; Non-empty-cell design; Strong orthogonal array; Addelman–Kempthorne construction
• ISSN: 0378-3758; 1873-1171
• DOI: 10.1016/j.jspi.2022.04.007

Strong orthogonal arrays are a class of space-filling designs suitable for computer experiments. Among this class of designs, those of strength two plus are most attractive as they aim at better space-filling properties in two-dimensions and hence have most economical run sizes. Most of previous work on strong orthogonal arrays of strength two plus focuses on the use of regular designs. In this paper, we develop a method of constructing space-filling designs using nonregular designs. Designs so constructed have very flexible run sizes compared to those constructed from regular designs. Apart from some theoretical results, computer searches are conducted to find space-filling designs using two-level nonregular designs of up to 40 runs and three-level nonregular designs of 27 and 54 runs. •Space-filling designs are constructed from nonregular designs.•Non-empty-cell designs are introduced and studied.•Theoretical constructions are investigated.•Design tables are presented for small run sizes.