本书介绍了多元数据分析的现代方法，主要讲解多元统计学中的最新方法及其应用。作者通过大量的示例说明每种技术的工作方式以及应用方法，还应用几何图形的方法来开发学生的直觉力，帮助读者对各种方法有一个比较形象的认识。书中大量习题和示例采用了来源于心理学，社会学和营销学等各个学科的真实数据。　
　　因为本书提供了各种类型的应用，所以适用于很多专业的教学，不仅适合营销学、组织行为学、会计学专业，还适合工程学、教育学、经济学、心理学、社会学和统计学等专业。

James M．Lattin于曼彻斯特理工大学获得博土学位，现为斯坦福大学商学院研究生院教授，自1984年以来一直在该校从事教学工作，主要教授营销管理和数据分析课程。他的主要研究方向是选择行为等数据库营
销。　　
　　J. Douglas Carroll于普林斯顿大学获得博土学位，现为罗格斯大学管理学院研究生院教授，他的主要研究方向是多维换算和数据分析技术，特别是其在营销学和心理学方面的应用。
　　Paul E．Green宾夕法尼亚大学教授，他的研究主要强调市场分析及客户研究中的量化方法和新的测量技术。

Once upon a time, over two decades ago now, two gentlemen (Paul Green and Doug Carroll) collaborated on a textbook titled Analyzing Multivariate Data. Their objec-tive was to produce a book with a pragmatic orientation--"a book for the data ana-lyzer." Quoting from the preface of that book,
　　Most users of multivariate statistical techniques are not professional statisticians. They are applications-oriented researchers--psychologists, sociologists, marketing research-ers, management scientists, and so on--who, from time to time, need the techniques to help them in their work. This text has been written for them and for students of these disciplines　....　As implied by the title, emphasis on data analysis and the objectives of people who do data analysis has shaped the character of the whole enterprise.
　　Many people adopted the book, including a young professor (Jim Lattin) who was teaching a course on multivariate data analysis for the very first time. The level of the text seemed quite appropriate for the mix of graduate students taking the course (mainly first- and some second-year graduate students from different parts of the uni-versity).　It was not too difficult (i.e., it did not rely too heavily on mathematics be-yond the preparation of the typical student) and not too simplistic (i.e., it was not a"cookbook,). Because the book presented a variety of applications, it appealed to a relatively broad cross-section of students (not only students in marketing, organiza-tional behavior, and accounting from the Graduate School of Business, but also stu-dents in engineering, education, economics, food research, psychology, sociology,
and statistics).
　　But perhaps the best feature of the book (in the opinion of the young professor) was the way the authors used the geometry underlying the mathematics to show how the techniques really worked. Even a student with only a tentative grasp of matrix al-gebra can see what is happening when he or she understands that each matrix opera-tion corresponds to a stretching (or shrinking) and rotation of the data. After the orig-inal text went out of print, the young professor continued to teach the course from the notes he had developed. Many things about the course changed (e.g., topics were added, dropped, and rearranged; new examples and larger data sets were included to keep pace with the increased computational capabilities of today's software pack-
ages), but the underlying pedagogy remained the same.
　　This new book is the result of the collaboration between the now not-so-young professor and the two authors of the original text. It is not so much a revision as it is a rebirth: a fresh look at multivariate techniques more than 20 years later, with new examples, new data, and some new methods, but grounded in the same pedagogical approach (applications-oriented, intuitively motivated using the underlying geometry of the method) that guided the creation of the original.
　　Organization The book is organized into three parts. By way of introduction, Part I (Chapters 1 through 3) provides a general overview of multivariate methods, some helpful back- ground on vectors and matrices and their geometrical interpretation, and a review of multiple regression analysis. Part II (Chapters 4 through 8) focuses on the analysis of interdependence, both among variables (principal components, factor analysis) and among objects (multidimensional scaling, cluster analysis). Part III (Chapters 9 through 13) covers canonical correlation and methods used in the analysis of depen- dence, including structural equation models with latent variables, logit choice mod- els, and special cases of the general linear model (analysis of variance, discriminant analysis).

Objectives
　　Our objective is to make students intelligent users of these multivariate techniques and good critics of multivariate analyses performed by others. If students are to be intelligent users and good critics of the techniques discussed in this book, they must have some grasp of theory, application, and interpretation. In other words, they must　 Have some intuition as to how the technique works. To this end, we use a geo- metric interpretation to provide the students with a mental picture of how each　method works. We use mathematics to support the underlying intuition (rather than as a substitute for it).
　　Be able to apply the technique. We take a hands-on approach, providing illus- trative examples in each chapter based on real-world data. To facilitate the ap- plication of these methods, we have developed student workbooks specific to

PART 1
Overview
1　Introduction
1.1　The Nature of Multivariate Data
1.2　Overview of Multivariate Methods
1.3　Format of Succeeding Chapters
2 Vectors and Matrices
2.1　Introduction
2.2　Definitions
2.3　Geometric Interpretation of Operations
2.4　Matrix Properties
2.5　Learning Summary
Exercises
3 Regression Analysis
3.1　Introduction
3.2　Regression Analysis: How It Works
3.3　Sample Problem: Leslie Salt Property
3.4　Questions Regarding the Application of Regression Analysis
3.5　Learning Summary
PART II
Analysis of Interdependence
4 Principal Components Analysis
4.1　Introduction
4.2　Principal Components: How It Works
4.3　Sample Problem: Gross State Product
4.4　Questions Regarding the Application of Principal Components
4.5　Learning Summary
5 Exploratory Factor Analysis
5.1　Introduction
5.2　Exploratory Factor Analysis: How It Works
5.3　Sample Problem: Perceptions of Ready-to-Eat Cereals
5.4　Questions Regarding the Application of Factor Analysis
5.5　Learning Summary
6 Confirmatory Factor Analysis
6.1　Introduction
6.2　Confirmatory Factor Analysis: How It Works
6.3　Sample Problems
6.4　Questions Regarding the Application of Confirmatory Factor Analysis
6.5　Learning Summary
7 Multidimensional Scaling
7.1　Introduction
7.2　Classical Metric MDS: How It Works
7.3　Nonmetric MDS: How It Works
7.4　The INDSCAL Model and Method for Individual Differences Scaling:
How It Works
7.5　Multidimensional Analysis of Preference: How It Works
7.6　Learning Summary
7.7　Selected Readings
8 Cluster Analysis
8.1　Introduction
8.2　Objectives of Cluster Analysis
8.3　Measures of Distance, Dissimilarity, and Density
8.4　Agglomerative Clustering: How It Works
8.5　Partitioning: How It Works
8.6　Sample Problem: Preference Segmentation
8.7　Questions Regarding the Application of Cluster Analysis
8.8　Learning Summary
PART III
Analysis of Dependence
9 Canonical Correlation
9.1　Introduction
9.2　Canonical Correlation: How It Works
9.3　Sample Problem
9.4　Questions Regarding the Application of Canonical Correlation
9.5　Learning Summary
10 Structural Equation Models with
Latent Variables
10.1　 Introduction
10.2　Structural Equation Models with Latent Variables: How It Works
10.3　Sample Problem: Modeling the Adoption of Innovation
10.4　Questions Regarding the Application of Structural Equations with
Latent Variables
10.5　Learning Summary
11　Analysis of Variance
11.1　Introduction
11.2　ANOVA/ANCOVA: How It Works
11.3　Sample Problem: Test Marketing a New Product
11.4　Multiple Analysis of Variance (MANOVA): How It Works
11.5　Sample Problem: Testing Advertising Message Strategy
11.6　Questions Regarding the Application of ANOVA
11.7　Learning Summary
12 Discriminant Analysis
12.1　Introduction
12.2　Two-Group Discriminant Analysis: How It Works
12.3　Sample Problem: Books by Mail
12.4　Questions Regarding the Application of Two-Group Discriminant Analysis
12.5　Multiple Discriminant Analysis: How It Works
12.6　Sample Problem: Real Estate
12.7　Questions Regarding the Application of Multiple Discriminant Analysis
12.8　Learning Summary
13 Logit Choice Models
13.1　Introduction
13.2　Binary Logit Model: How It Works
13.3　Sample Problem: Books by Mail
13.4　Multinomial Logit Model: How It Works
13.5　Sample Problem: Brand Choice
13.6　Questions Regarding the Application of Logit Choice Models
13.7　Learning Summary
Statistical Tables
Bibliography
Index
particular statistical packages (e.g., SAS and SPSS). These workbooks explain
how the concepts in the text are linked to the application software and show the
student how to perform the analyses presented in each chapter. The program
templates provided in the workbooks enable students to run their own analyses
of the more than 100 data sets (most taken from real applications in the pub-
lished literature) contained the CD-ROM that accompanies the text.
Be able to interpret the results of the analysis. In each chapter, we raise the im-
portant issues and problems that tend to come up with the application of each
method. We place special emphasis on assessing the generalizability of the re-
sults of an analysis, and suggest ways in which students can test the validity of
their findings.