ComputerGenerated Minimal (and Larger) ResponseSurface Designs.pdfVIP

Computer-GeneratedMinimal(andLarger)Response-Surface Designs: (II)TheCube R.H.Hardinand N.J.A.Sloane1 Mathematical Sciences Research Center ATT Bell Laboratories 600 Mountain Avenue Murray Hill, New Jersey 07974 This paper and its companion (Part I) were written in 1991 but never published. They are now (August, 2001) being published electronically on N. J. A. Sloane’s home page, /∼njas/. ABSTRACT Computer-generated designs in the cube are described which have the minimal (or larger) number of runs for a full quadratic response-surface design. Examples of 2-factor designs are included with 6 to 20 runs, 3-factor designs with 10 to 20 runs, 4-factor designs with 15 to 20 runs, 5-factor designs with 21 to 25 runs, 6-factor designs with 28 to 31 runs, and 7-factor designs with 36 and 39 runs. The designs were constructed by minimizing the average prediction variance, and without imposing any prior constraints – such as a central composite structure – on the locations of the points. Key Words. Minimal designs; cube designs; quadratic response surface; computer-generated designs; minimal variance designs. 1Presentaddress:ATTShannonLabs,FlorhamPark,NJ07932-0971 1 Introduction In Part I (Hardin and Sloane, 1991) we constructed designs for a quadratic response- surface model in which the runs are made at points in or on a sphere. The present paper uses the same techniques to construct designs in which the runs are at points (x ,...,x ) 1 k in or on the cube defined by − 1 ≤