The use of second-order ‘spherical’ and ‘cuboldal’ designs In the wrong regions

Abstract

Regions of interest in the space of experimental factors to be examined are usually determined before experimentation begins, and can be defined in many ways. Two basic region shapes, in coded variable co-ordinate systems, are the unit sphere and the cuboid ( ± 1, ± 1,…,± 1), and suitable designs for these are known. This paper examines the consequences of using designs appropriate for a spherical region in a cuboidal region, and vice versa.