Enhanced descriptions from Syndetics:

Multivariate analysis is an important tool for social researchers, but the subject is broad and can be quite technical for those with limited mathematical and statistical backgrounds. To effectively acquire the tools and techniques they need to interpret multivariate data, social science students need clear explanations, a minimum of mathematical detail, and a wide range of exercises and worked examples.

Classroom tested for more than 10 years, The Analysis and Interpretation of Multivariate Data for Social Scientists describes and illustrates methods of multivariate data analysis important to the social sciences. The authors focus on interpreting the pattern of relationships among many variables rather than establishing causal linkages, and rely heavily on numerical examples, visualization, and on verbal , rather than mathematical exposition. They present methods for categorical variables alongside the more familiar method for continuous variables and place particular emphasis on latent variable techniques.

Ideal for introductory, senior undergraduate and graduate-level courses in multivariate analysis for social science students, this book combines depth of understanding and insight with the practical details of how to carry out and interpret multivariate analyses on real data. It gives them a solid understanding of the most commonly used multivariate methods and the knowledge and tools to implement them.

Datasets, the SPSS syntax and code used in the examples, and software for performing latent variable modelling are available at http://www.mlwin.com/team/aimdss.html>

Table of contents provided by Syndetics

• Preface (p. ix)
• 1 Setting the Scene (p. 1)
• 1.1 Structure of the book (p. 1)
• 1.2 Our limited use of mathematics (p. 4)
• 1.3 Variables (p. 7)
• 1.4 The geometry of multivariate analysis (p. 9)
• 1.5 Use of examples (p. 10)
• 1.6 Data inspection, transformations, and missing data (p. 12)
• 1.7 A final word (p. 13)
• 1.8 Reading (p. 13)
• 2 Cluster Analysis (p. 15)
• 2.1 Classification in social sciences (p. 15)
• 2.2 Some methods of cluster analysis (p. 18)
• 2.3 Graphical presentation of results (p. 23)
• 2.4 Derivation of the distance matrix (p. 27)
• 2.5 Example on English dialects (p. 31)
• 2.6 Comparisons (p. 37)
• 2.7 Clustering variables (p. 39)
• 2.8 Further examples and suggestions for further work (p. 39)
• 2.9 Further reading (p. 51)
• 3 Multidimensional Scaling (p. 53)
• 3.1 Introduction (p. 53)
• 3.2 Examples (p. 55)
• 3.3 Classical, ordinal, and metrical multidimensional scaling (p. 59)
• 3.4 Comments on computational procedures (p. 62)
• 3.5 Assessing fit and choosing the number of dimensions (p. 63)
• 3.6 A worked example: dimensions of colour vision (p. 64)
• 3.7 Further examples and suggestions for further work (p. 68)
• 3.8 Further reading (p. 79)
• 4 Correspondence Analysis (p. 81)
• 4.1 Aims of correspondence analysis (p. 81)
• 4.2 Carrying out a correspondence analysis: a simple numerical example (p. 83)
• 4.3 Carrying out a correspondence analysis: the general method (p. 88)
• 4.4 The biplot (p. 91)
• 4.5 Interpretation of dimensions (p. 95)
• 4.6 Choosing the number of dimensions (p. 97)
• 4.7 Example: confidence in purchasing from European Community countries (p. 99)
• 4.8 Correspondence analysis of multi-way tables (p. 105)
• 4.9 Further examples and suggestions for further work (p. 109)
• 4.10 Further reading (p. 114)
• 5 Principal Components Analysis (p. 115)
• 5.1 Introduction (p. 115)
• 5.2 Some potential applications (p. 116)
• 5.3 Illustration of PCA for two variables (p. 117)
• 5.4 An outline of PCA (p. 120)
• 5.5 Examples (p. 122)
• 5.6 Component scores (p. 129)
• 5.7 The link between PCA and multidimensional scaling, and between PCA and correspondence analysis (p. 132)
• 5.8 Using principal component scores to replace original variables (p. 134)
• 5.9 Further examples and suggestions for further work (p. 136)
• 5.10 Further reading (p. 141)
• 6 Factor Analysis (p. 143)
• 6.1 Introduction to latent variable models (p. 143)
• 6.2 The linear single-factor model (p. 146)
• 6.3 The general linear factor model (p. 149)
• 6.4 Interpretation (p. 152)
• 6.5 Adequacy of the model and choice of the number of factors (p. 154)
• 6.6 Rotation (p. 156)
• 6.7 Factor scores (p. 160)
• 6.8 A worked example: the test anxiety inventory (p. 161)
• 6.9 How rotation helps interpretation (p. 166)
• 6.10 A comparison of factor analysis and principal component analysis (p. 167)
• 6.11 Further examples and suggestions for further work (p. 169)
• 6.12 Further reading (p. 174)
• 7 Factor Analysis for Binary Data (p. 175)
• 7.1 Latent trait models (p. 175)
• 7.2 Why is the factor analysis model for metrical variables invalid for binary responses? (p. 178)
• 7.3 Factor model for binary data (p. 179)
• 7.4 Goodness-of-fit (p. 184)
• 7.5 Factor scores (p. 188)
• 7.6 Rotation (p. 190)
• 7.7 Underlying variable approach (p. 190)
• 7.8 Example: sexual attitudes (p. 192)
• 7.9 Further examples and suggestions for further work (p. 197)
• 7.10 Software (p. 206)
• 7.11 Further reading (p. 206)
• 8 Factor Analysis for Ordered Categorical Variables (p. 207)
• 8.1 The practical background (p. 207)
• 8.2 Two approaches to modelling ordered categorical data (p. 208)
• 8.3 Item response function approach (p. 209)
• 8.4 Examples (p. 216)
• 8.5 The underlying variable approach (p. 219)
• 8.6 Unordered and partially ordered observed variables (p. 224)
• 8.7 Further examples and suggestions for further work (p. 228)
• 8.8 Software (p. 234)
• 8.9 Further reading (p. 234)
• 9 Latent Class Analysis for Binary Data (p. 235)
• 9.1 Introduction (p. 235)
• 9.2 The latent class model for binary data (p. 236)
• 9.3 Example: attitude to science and technology data (p. 241)
• 9.4 How can we distinguish the latent class model from the latent trait model? (p. 245)
• 9.5 Latent class analysis, cluster analysis, and latent profile analysis (p. 247)
• 9.6 Further examples and suggestions for further work (p. 248)
• 9.7 Software (p. 252)
• 9.8 Further reading (p. 252)
• References (p. 253)
• Index (p. 257)

Reviews provided by Syndetics

CHOICE Review

Bartholomew, Steele, Moustaki, and Galbraith clearly present their ideas by relying heavily on numerous numerical examples and on verbal rather than mathematical arguments. The contents (nine chapters) include some basic ideas useful for handling multivariate data. Chapters treat methods such as cluster analysis, multidimensional scaling, correspondence analysis, and principal components analysis for describing, summarizing, and exploring multivariate data sets. The last four chapters include model-based methods, leading readers to make inferences about processes that have generated data; these methods are factor analysis, factor analysis for binary data, factor analysis for ordered categorical variables, and latent class analysis for binary data. Computations throughout can be easily handled by a computer with standard software provided at the publisher's Web site . The authors interpret the pattern of relationships among several variables rather than establishing causal linkages, and place particular emphasis on latent variable techniques in their presentation. The authors' easy style makes reading this book a great pleasure. A drawback is the absence of problem sets for students. Chapter references; comprehensive bibliography; combined author-subject index; list of symbols used in the book. A very useful and interesting book, excellent for researchers and students for analyzing and interpreting multivariate data in social sciences. Upper-division undergraduates through professionals. D. V. Chopra Wichita State University

Author notes provided by Syndetics

David J. Bartholomew is Emeritus Professor of Statistics at the London School of Economics and Political Science
Fiona Steele is a Research Lecturer in Statistics at the Institute of Education, University of London
Irini Moustaki is a Lecturer in Social Statistics and Jane I. Galbraith is a Senior Teaching Fellow in Statistics at the London School of Economics and Political Science