Applied Longitudinal Analysis
暂无分享,去创建一个
dents and as a reference for those researchers embarking on running their own designed experiment, provides thorough coverage of a wide variety of topics for a number of experimental situations. The book assumes familiarity with basic statistical methods such as sampling, multiple linear regression, simple tests of hypotheses, confidence intervals, and analysis of variance, for which background can be readily found from a wide variety of introductory statistics books. The book’s strength is found in the breadth of its appeal. It begins with good coverage of the basics, such as completely randomized designs, linear models, hypothesis testing, power, and blocking, which will help those who have not had much exposure to design of experiments gain confidence with the basic principles and goals of a well-designed study. In addition to this clear introduction to the basics, the book examines a broad range of design types, focused almost exclusively on designs for variables with categorical levels and a corresponding analysis of variance. With a primary emphasis on design, not analysis, the book includes many details about a breadth of design types but does not place a strong emphasis on analysis or on finding optimal combinations of factors through multiple comparisons or other approaches. With an accompanying website (http://www4.stat.ncsu.edu/ ̃gumpertz/ ggdata) to provide electronic copies of the data and SAS code for examples and exercises in the book, using the book as a graduate textbook is straightforward, and most students will find the book readable and helpful for grasping the important ideas. Because there is much more material in the book than can be covered in a oneor two-semester course on design of experiments, the instructor should have flexibility to focus the course on a subset of topics most relevant to the class. A handful of exercises at the end of each chapter involve both design and analysis for applications in agricultural, engineering, and biological applications. For a practitioner, the well-organized format and style of presentation makes finding and understanding the relevant chapter easy. Chapters 1–5 provide an introduction to design of experiments, completely randomized designs, linear models, hypothesis testing, choice of sample size appropriate for desired power, and the advantages of blocking. These chapters are clearly written and go through the basics in a comfortable and nonnotationally rich approach. Throughout the book, the authors provide alternative notations for describing designs, which will make using this book in conjunction with other resources more straightforward. Although the ordering of the more advanced topics in the remainder of the book is a bit nonstandard, a vast collection of possible designs is described. Chapter 6 looks at latin squares, greco-latin squares, and variations of these designs. Chapter 7 gives a nice overview of split-plot and strip-plot designs with further variations, including how covariates can be incorporated in these designs. Chapter 8 and 9 consider incomplete-block designs and repeatedtreatment designs. Chapters 10 and 11 deal with 2 and 3 factorial designs. The short Chapter 12 discusses analysis techniques for experiments without a measure of pure error. Chapters 13 and 14 build on the basics of factorial designs presented in Chapters 10 and 11, with discussion of blocking and confounding issues for these cases, fractional factorial designs, and selecting between different aliasing structures. Chapter 15 is one of the few chapters that considers continuous factor levels through the basics of response surface designs. It would be beyond the scope of this book to provide a more detailed coverage of the material; if this were an area of interest, then Myers and Montgomery (2002) would likely be the most appropriate reference. Chapter 16 presents Plackett–Burman Hadamard designs, and Chapter 17 considers p and nonstandard factorial designs. Chapter 18 discusses designs where run order is important (or restricted), and Chapter 19 examines sequences of fractions of factorial designs that involve foldover techniques or augmentation. Chapters 20–25 discuss an assortment of other designs, including factorial experiments with quantitative factors, supersaturated plans, multistage experiments, orthogonal arrays, and computer experiments. Each of these chapters provides an introduction to broad areas. The authors should be commended for their inclusion of these topics into a graduate-level text and for their clear presentation of the basics. The book does an excellent job showing the reader how to construct a design of a given type for a wide variety of factors and levels and providing the mechanics of the analysis to determine if factors have an effect. The authors could have done a better job developing how to take the analysis one step further to determine the nature of relationships, how to optimize the process, or how to draw conclusions about what has been observed, but perhaps this is just too big a mandate for a single book. Given the wide variety of choices presented throughout the book, there is relatively little on how to choose between competing categories of designs for a particular situation. Including a section on design assessment through formal optimality criteria or subjective measures would have been a welcome addition, and would have helped the practitioner designing an experiment to make judicious choices and better understand the trade-offs between designs. For a practitioner suddenly faced with a very large collection of tools from which to choose, little intuition or insight is provided about how to select a design most appropriate for a given number of factors, with restrictions on how the data can be collected. Finally, I think the authors might have emphasized how the different designs naturally put limitations on what types of models are possible and what types of relationships between factors and responses can be considered. Because considering such issues is a natural starting point for choosing a design, these are important shortcomings. Overall, this book is an excellent reference for statisticians and practitioners who would like to gain broad exposure to the tools available for studying relationships between qualitative and quantitative factors and the observed responses. The book does a nice job of discussing the designs in a context of a planning phase, an execution phase, and an analysis phase. It also includes an extensive list of references to guide the reader to supplementary materials and literature on the topics for which only an introduction could be provided.
[1] Stan Lipovetsky,et al. Generalized Latent Variable Modeling: Multilevel,Longitudinal, and Structural Equation Models , 2005, Technometrics.