ABSTRACT
The D-criterion is a popular optimality criterion for computer-generated screening designs. Because hypothesis testing for the evaluation of statistically significant effects is one of the focal points of variable screening, the D-criterion is an intuitive choice due to its focus upon precise parameter estimation. In many situations, however, there is not a single unique set of points that comprises a D-optimal design. The designs in the D-optimal set often perform quite differently in terms of other important criteria. In this manuscript we consider a suite of optimality criteria for consideration when choosing a good screening design.
Notes
Best values for each column are in bold; worst values are in italics.