Elfving's theorem for \(R\)-optimality of experimental designs (Q2174526)

The authors consider the criterion of \(R\)-optimality for experimental designs in a multiresponse linear model. The \(R\)-optimality criterion proposed by \textit{H. Dette} [J. R. Stat. Soc., Ser. B 59, No. 1, 97--110 (1997; Zbl 0884.62081)] minimizes the product of the diagonal elements of the information matrix. This is proportional to minimization of the volume of the Bonferroni rectangular confidence region for the regression parameters. In this paper the authors prove an Elfving's theorem for \(R\)-optimality. Elfving's theorems are very helpful in optimal design theory because their elegant geometrical characterization of optimality. To illustrate this, the authors present four examples of construction of \(R\)-optimal designs.