Analysis of Multivariate Social Science Data: Statistical Machine Learning Methods

Drawing on the authors' varied experiences researching and teaching in the field, Analysis of Multivariate Social Science Data: Statistical Machine Learning Methods, Third Edition, enables a basic understanding of how to use key multivariate methods in the social sciences. With minimal mathematical and statistical knowledge required, this third edition expands its topics to include graphical modelling, models for longitudinal data, structural equation models for categorical variables, and latent class analysis for ordinal, nominal and continuous variables. It also connects the topics to terminology and principles of machine learning, intended to help readers grasp the links between methods of multivariate analysis and advancements in the field of data science.

After describing methods for the summarisation of data in the first part of the book, the authors consider regression analysis. This chapter provides a link between the two halves of the book, signalling the move from descriptive to inferential methods. The remainder of the text deals with model-based methods that primarily make inferences about processes that generate data.

Relying heavily on numerical examples from a range of disciplines, the authors provide insight into the purpose and working of the methods as well as the interpretation of results from analyses. Many of the same examples are used throughout to illustrate connections between the methods. In most chapters, the authors present suggestions for further work that go beyond conventional practice, encouraging readers to explore new ground in social science research.

Features

• Contains new chapters on undirected graphical modelling, and models for longitudinal data, as well as new material such as K-means, cross-validation, structural equation models for categorical variables, latent class analysis for categorical, nominal and continuous variables and treatment of missing data.
• Connects topics with terminology and principles of machine learning.
• Presents numerous examples of real-world applications, including voting preferences, social attitudes, educational assessment, recidivism and health.
• Covers methods that summarise, describe, and explore multivariate datasets, including longitudinal data.
• Establishes a unified approach to latent variable modelling by providing detailed coverage of methods such as item response theory, factor analysis for continuous and categorical data and models for categorical latent variables.
• Covers models for hierarchical and longitudinal data and their connection to latent variable models.
• Offers a full version of the data sets in the text or the book's website, with software code for implementing the analyses on the website.

The book offers a balanced and accessible resource for students and researchers with limited mathematical and statistical training. It serves as a practical resource for courses in multivariate analysis and as a guide for applying these techniques in applied research.

Irini Moustaki is a professor of Statistics in the Department of Statistics at the London School of Economics and Political Science. She received her bachelor's degree in Statistics and Computer Science from the Athens University of Economics and Business and her MSc and PhD in Statistics from the LSE. Her research interests are in latent variable models and structural equation models. Her methodological work includes treating missing data, longitudinal data, outlier detection, goodness-of-fit tests, and advanced estimation methods. Furthermore, she has made methodological and applied contributions to comparative cross-national studies and epidemiological studies on rare diseases. Irini received an honorary doctorate from the Faculty of Social Sciences, Uppsala University, in 2014. She is a Fellow of the British Academy. She was the Executive Editor of the journal Psychometrika from November 2014 to December 2018 and the President of the Psychometric Society from July 2021 to July 2022.

Fiona Steele is a Professor of Statistics in the Department of Statistics at the London School of Economics and Political Science (LSE). She holds a Ph.D. in Social Statistics from the University of Southampton. Her research interests are in developments of statistical methods that are motivated by social science problems. Her areas of expertise include longitudinal data analysis, multilevel and latent variable modelling, and dyadic data analysis. She has worked on a range of applications in demography, education, family psychology and health. Fiona has directed several research grants on methods for multilevel and longitudinal data analysis. She also led the development of, and contributed modules to, the popular online 'LEMMA' course on multilevel modelling. Fiona is a Fellow of the British Academy and was awarded a CBE and the Royal Statistical Society Howard Medal for her contributions to social statistics.

Yunxiao Chen is an Associate Professor of Statistics in the Department of Statistics at the London School of Economics and Political Science (LSE). He holds a Ph.D. in Statistics from Columbia University in the City of New York. His research focuses on the intersection of multivariate statistics and machine learning, where he develops models, computational algorithms, and statistical theories for learning from complex data and applies them to education, psychology, and other social science disciplines. Dr. Chen has received numerous awards, including the 2018 Brenda H. Lloyd Dissertation Award from the National Council on Measurement in Education and the 2022 Early Career Award from the Psychometric Society. He was also a Spencer Foundation/NAEd Postdoctoral Fellow at the United States National Academy of Education from 2018 to 2020. His work has appeared in leading journals in statistics and machine learning, such as the Journal of the American Statistical Association, Biometrika, Journal of the Royal Statistical Society, Series A (Statistics in Society), and the Journal of Machine Learning Research. Additionally, Dr. Chen serves as an associate editor for several prominent publications, including Psychometrika, the British Journal of Mathematical and Statistical Psychology, the Journal of Educational and Behavioural Statistics, and Psychological Methods. This book, Analysis of Multivariate Social Science Data: Statistical Machine Learning Methods, draws on his years of experience teaching and researching in the field.

David John Bartholomew was a professor of Statistics in the Department of Statistics at the London School of Economics and Political Science from 1973 to 1996, when he became an emeritus professor. His research interests were in the areas of stochastic modelling, social measurement, factor analysis and latent variable modelling. Bartholomew published some 25 books and more than 140 research papers. He served as pro-director of the LSE from 1998 to 1991. He served as co-editor of the Journal of the Royal Statistical Society, Series B, from 1966 to 1969 and president from 1993 to 1995. He was awarded the Guy medal in bronze in 1971. He was elected a fellow of the British Academy in 1987. Professor Bartholomew passed away in 2017.