This monograph collects the author’s research on several topics in the analysis of multiattribute data. It consists of two parts which contain nine chapters in all. As the author said, “multiattribute data appear in studies about, or using multiattribute preference models”, these models have been useful in many disciplines. Many authors have made important contributions in this area in the past several decades. The author’s research in this monograph mainly concentrates on model analysis under various meaningful restrictions by implementing the ideas and techniques in nonlinear multivariate analysis (NMVA for short). Exploration on algorithms for multiattribute preference models (MPM for short) under some particular restrictions and Monte Carlo studies also constitute two important parts of the contents.
The first part of the monograph is mainly concerned with the individual-level analysis of hybrid MPM’s. Starting with an introduction to MPM and MPM data and their connection with NMVA in the first two chapters, the author finds a good way to lead the reader into the scope which will be covered in the research. After reading the morals in Chapter 2, the reader who has a basic knowledge in statistics will be easily accessible to the main ideas of the research. Based on the foundations laid in the first two introductory chapters, the author focuses on the individual-level analysis of hybrid MPM’s (Chapter 3 and Chapter 4), which constitute the main contents of the first part of the monograph. As an important type of compositional MPM’S, expectancy-value type models, which are studied in Chapter 3, are easy-to-understand linear models. The reader with a basic knowledge in this subject will find it easy to catch up with the idea of the two basic purposes of NMVA: estimating model parameters and scaling parameters according to some structures of restrictions on the parameters. By formulating complex restrictions on unknown parameters in terms of the intersection of a limited number of closed convex cones, the author makes the algorithms to be easily realized in computer programs.
The second part of the monograph is composed of three chapters. Several topics related to the vector model are investigated in Chapter 6. By the vector model, the models in the first part of the monograph can be easily interpreted. Techniques for fitting the vector model and some relations to other models and techniques are explored. Comparability is another kind of restrictions on observations. Chapter 7 presents the author’s research on various issues in this topic. The introduction of the concept “multiple principal registers” is an important contribution to model individual differences in quantifications of comparable observations. The idea in principal component analysis finds its similarity here. To illustrate the theoretical developments in the second part, the author made some empirical study in Chapter 8, which helps the reader to understand well the principles and techniques developed in the second part. As a concluding and the last chapter of the monograph, in Chapter 9 the author points out that further research on the MPM’s seems promising and the proposed MPM’s and techniques may find more applications in other disciplines.
As a monograph in the field of applied sciences, this book is a good example in applications of techniques in nonlinear multivariate analysis. Readers majored in the fields such as social and behavioral sciences, marketing research and applied statistics will find it useful to make their attempt in dealing with similar problems by the techniques in this monograph. It seems that the techniques introduced in this monograph would be more easily acquired by the reader if the author had paid more attention to the background of the models and its various restrictions and to the illustrations of the techniques.