Optimal experimental design

Optimal experimental design is a class of designs, within the wider practice of design of experiments, that are best with respect to some statistical criterion. The creation of the field is credited to the Danish statistician Kirstine Smith. When statistical models are estimated, an optimal design lets the parameters be estimated without bias and with the smallest possible variance, whereas a design that is not optimal needs more experimental runs to reach the same precision, so a good design can lower the cost of experimentation in practical terms. How good a design is depends on the statistical model and is judged against a criterion related to the variance matrix of the estimator. Choosing a suitable model and a suitable criterion function both call for an understanding of statistical theory together with practical experience of designing experiments.