Functional approach to optimal experimental design. (Q820225)

``The book presents a novel approach for studying optimal experimental designs'' (from the cover). The ``functional approach'' proposed here is based on an extensive usage of the implicit function theorem which allows for approximate constructions of optimal designs in linear and nonlinear regression setups. Chapters~1 and 2 provide the fundamentals of the theory of optimal designs and the proposed functional approach with first applications to a particular class of nonlinear exponential models and a three-parameter logistic regression model. Locally \(D\)-optimal and maximin \(D\)-efficient designs are considered with minimal support for which the optimization problem simplifies: As the optimal weights are all equal only the optimal supporting points of the design have to be determined. In the subsequent chapters the functional approach is applied to various types of regression models: Chapters~3 and 4 deal with linear regression models. In Chapter~3 optimal designs are obtained for polynomial models with respect to the maximum eigenvalue (\(E\)-) and the \(e_k\)-criterion for the leading term of the polynomial while Chapter~4 provides \(D\)- and \(E\)-optimal designs for trigonometric regression models. The remaining chapters are devoted to a collection of results for particular nonlinear regression models and varying design criteria: locally \(D\)-optimal designs are presented for rational models where the mean response is given by the ratio of two polynomials in Chapter~5 and for a more general exponential model with additional polynomial terms in Chapter~6. For those two models \(E\)- and certain \(c\)-optimal designs are obtained in Chapter~7. The final Chapter~8 is devoted to the Monod model where the mean response is implicitly defined as the solution of a differential equation. For the Monod model locally \(D\)-, \(E\)- and \(e_k\)-optimal as well as maximin efficient designs are characterized. The findings are illustrated by numerical calculations. This book is essentially a collection of already published research results of its author which are thereby made available to a broader audience. However, also some new material is included like the Taylor expansions for the Monod model in section 8.5.