scientific article; zbMATH DE number 1193553

MaRDI QIDQ3842004zbMATH OpenFDO

Authors

Publication date 20 August 1998

balanced incomplete block designs confounding fractional factorial experiments random effects model response surface designs split-plot designs

Mathematics Subject Classification ID

Combinatorial aspects of block designs (05B05) Orthogonal arrays, Latin squares, Room squares (05B15) Introductory exposition (textbooks, tutorial papers, etc.) pertaining to statistics (62-01) Statistical block designs (62K10) Factorial statistical designs (62K15) Design of statistical experiments (62K99) Design of statistical experiments (62Kxx)

Review text

This book on Statistical Design and Analysis of Experiments is one of several important books that were allowed to go out of print and are now being republished as part of the series titled ``Classics in applied mathematics by the Society for Industrial and Applied Mathematics (SIAM). The book was first published in 1971 by Macmillan Company, see the review Zbl 0231.62089. It is aimed at mathematically oriented readers and is suitable as a text book at the graduate level. A working knowledge of matrix algebra and prior exposure to mathematical statistics are the assumed prerequisites. In addition to the original preface, this unabridged reprinting of the original version has a new preface written by the author specifically for the Classics Edition where he shares with the readers some of the history behind the subject along with his thoughts on how computers have influenced the field of design. He also discusses topics he would add and topics he would modify or omit if he were to write such a book today. This preface, though short, provides much insight, and anyone interested in the subject will find it extremely interesting. There is also a section at the end of this preface where a complete list of references mentioned in the preface is given. The book itself has fifteen chapters and an appendix on matrices and quadratic forms. Chapter 1 provides an overview of topics that will be addressed in the rest of the book. Chapter 2 sets up the linear model using matrix notation. Normal equations are derived and their solutions are discussed for the full rank case as well as for the singular case. Important results concerning the distributions of quadratic forms and independence of quadratic forms, in normal variables, are given. Chapters 3 through 12 deal with the most commonly used experimental design. The single factor experiment is introduced in Chapter 3. Analysis of data from such experiments is discussed both for the fixed effects model and the random effects model. Other topics dealt with in this chapter include the power of the \(F\)-test, the idea of blocking, the randomized complete block experiment, and the analysis of covariance. Experiments with two factors, with replication and without replication, are considered in Chapter 4. The random and the mixed effects models are also discussed here. Multifactor experiments form the topics of Chapter 5. The concepts of crossing and nesting of factors are introduced. General rules are given for calculating sums of squares and expected mean squares (EMS) in balanced experiments. The chapter concludes with a treatment of split-plot designs and split-split plot designs. Chapter 6 considers the Latin square design and its generalizations. These include the Graeco-Latin Square designs and sets of orthogonal Latin squares. Change-over designs with or without residual effects are discussed. Confounding and partial confounding in two and three level factorial experiments are examined in detail in Chapter 7. Confounding in more general factorial experiments is also examined here. Fractional factorial designs for two level experiments is the topic of Chapter 8. Some commonly used fractions are discussed and blocking schemes for fractional factorial experiments are examined. Analysis of data from standard fractional factorial experiments is discussed. Special consideration is given to resolution III, resolution IV, and resolution V designs. The idea of three-quarter replicates is introduced and some useful designs are given. Fractional factorial designs for factors with more than two levels are considered in Chapter 9. A discussion of Plackett-Burman designs and Adelman designs is provided. Some fractional designs are proposed for the mixed factorial experiment where some factors have two levels and others have three levels. Chapter 10 deals with response surface designs including first order designs, second order designs, polygonal designs and central composite designs. Fitting of response surfaces, search for optimum factor combinations, and the application of evolutionary operations are also discussed. Chapters 11 and 12 are devoted to a discussion of incomplete block designs. Chapter 11 considers balanced incomplete block designs, intra-block and inter-block estimates, recovery of inter-block information, combining intrablock and interblock estimates and other related topics. Chapter 12 deals with partially balanced incomplete block designs. Various classical association schemes are introduced and a discussion of how to select an incomplete block design is provided. Chapters 13 through 15 consider theoretical issues regarding existence and construction of balanced incomplete block designs, partially balanced incomplete block designs, and properties of various association schemes. These topics are accessible to the more mathematically trained reader. The book concludes with an appendix on matrices and quadratic forms. A bibliography and an index are provided. Needless to say, this book has been and will continue to be, an important book on the subject of experimental design. It will serve as a useful reference to statisticians and practitioners alike for many more years.

Cited in

(26)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3842004)