Optimum designs for multi-factor models (Q1921007): Difference between revisions

This book essentially reflects results of the Berlin research group on optimum designs. The main topic is the study of problems where different factors (e.g. qualitative and quantitative) may appear together in a linear model and of the effects which this has on optimum designs. In the book stress is rather on the discussion of the underlying structures with the aim to specify those structures which allow to use known designs of certain components to generate optimum designs for the whole model. The main technique is conditioning the model on each of its factors and the use of product designs. Various types of conditionings and a broad class of models illustrate this technique in part II of the book (with partial hints on its limitations); the classical \(D\)-, \(A\)- and \(E\)-criteria serve here as optimality criteria. These and all other necessary preparations (linear model, equivalence theorems, orthogonality and invariance) are presented in a very concise manner in the shorter introductory part I. As regards designs, the author restricts himself on approximate designs with the consequence that e.g. BIBDs or Hadamard matrices do not appear; they degenerate to uniform designs. Good knowledge of linear models and matrix calculus is indispensible for reading this book.