An update of optimal experimental designs for response surface methodology
Johnson, Robert W., writer on mathematical statistics
MetadataShow full item record
Many articles, reviews, and books on the subject of optimal designs for determining response surfaces have been published in the three or so decades since such designs were introduced. However, some important developments have occurred in the past five years. One of these is that l-optimal designs can now be easily obtained by one or more computer programs. These have two major advantages over other kinds of designs: 1. The criterion for optimality is closest to that usually described as most desirable; 2. One can usually obtain an l-optimal design which has quality features equivalent to alternates, but having fewer runs. Obviously, this feature takes on added importance when runs are very expensive and/or time-consuming. A second important development is the appearance of new or improved software tailored for a “userfriendly” approach to using response surface methodology. One such package has a module that enables an experimenter to compare alternate choices of designs on the basis of the ability to minimize the standard error of prediction in the experimental region of interest. This allows an experimenter, even one not highly trained in statistics, to make an intelligent choice before investing time and money in doing the experiments.