This book demonstrates how to use multilevel and longitudinal modeling techniques available in the IBM SPSS mixed-effects program (MIXED). Annotated screen shots provide readers with a step-by-step understanding of each technique and navigating the program. Readers learn how to set up, run, and interpret a variety of models. Diagnostic tools, data management issues, and related graphics are introduced throughout. Annotated syntax is also available for those who prefer this approach. Extended examples illustrate the logic of model development to show readers the rationale of the research questions and the steps around which the analyses are structured. The data used in the text and syntax examples are available at www.routledge.com/9780415817110.
Highlights of the new edition include:
- Updated throughout to reflect IBM SPSS Version 21.
- Further coverage of growth trajectories, coding time-related variables, covariance structures, individual change and longitudinal experimental designs (Ch.5).
- Extended discussion of other types of research designs for examining change (e.g., regression discontinuity, quasi-experimental) over time (Ch.6).
- New examples specifying multiple latent constructs and parallel growth processes (Ch. 7).
- Discussion of alternatives for dealing with missing data and the use of sample weights within multilevel data structures (Ch.1).
The book opens with the conceptual and methodological issues associated with multilevel and longitudinal modeling, followed by a discussion of SPSS data management techniques which facilitate working with multilevel, longitudinal, and cross-classified data sets. Chapters 3 and 4 introduce the basics of multilevel modeling: developing a multilevel model, interpreting output, and trouble-shooting common programming and modeling problems. Models for investigating individual and organizational change are presented in chapters 5 and 6, followed by models with multivariate outcomes in chapter 7. Chapter 8 provides an illustration of multilevel models with cross-classified data structures. The book concludes with ways to expand on the various multilevel and longitudinal modeling techniques and issues when conducting multilevel analyses. It's ideal for courses on multilevel and longitudinal modeling, multivariate statistics, and research design taught in education, psychology, business, and sociology.
Table of Contents
1. Introduction to Multilevel Modeling with IBM SPSS. 2. Preparing and Examining the Data for Multilevel Analyses.
3. Defining a Basic Two-Level Multilevel Regression Model. 4. Three-Level Univariate Regression Models. 5. Examining Individual Change with Repeated Measures Data. 6. Applications of Mixed Models for Longitudinal Data.
7. Multivariate Multilevel Models. 8. Cross-Classified Multilevel Models. 9. Concluding Thoughts. Appendix A: Syntax Statements. Appendix B: Model Comparisons Across Software Applications. Appendix C: Syntax Routine to Estimate Rho from Model’s Variance Components.
"Ronald Heck and his colleagues have provided academics, graduate students, and practitioners with a resource that few can surpass. This book contains excellent details for users with varying degrees of proficiency in multilevel modelling. It should be on the book shelf of anyone who claims to use this technique." – Timothy Teo, University of Auckland, New Zealand
"This book serves, not only as an introduction to using IBM SPSS for multilevel models, but as a wonderful introduction to multilevel models through empirical example. It is a wonderful resource for an undergraduate or graduate course on multilevel modeling." – Kevin Grimm, University of California, Davis, USA
"This book is ideal for individuals interested in learning about how to analyze different types of multilevel and longitudinal models using the MIXED procedure in IBM SPPS. The book methodically progresses from the simplest of models and designs to the more advanced ones. The presentation of statistical concepts is easy to follow, the data analysis examples are excellent, and the screen shots of the scripts and outputted results are thoroughly and effectively annotated." – George A. Marcoulides, University of California, Riverside, USA