Principles and Practice of Structural Equation Modeling

Description

Significantly revised, the fifth edition of the most complete, accessible text now covers all three approaches to structural equation modeling (SEM)–covariance-based SEM, nonparametric SEM (Pearl’s structural causal model), and composite SEM (partial least squares path modeling). With increased emphasis on freely available software tools such as the R lavaan package, the text uses data examples from multiple disciplines to provide a comprehensive understanding of all phases of SEM–what to know, best practices, and pitfalls to avoid. It includes exercises with answers, rules to remember, topic boxes, and a new self-test on significance testing, regression, and psychometrics. The companion website supplies helpful primers on these topics as well as data, syntax, and output for the book's examples, in files that can be opened with any basic text editor.
 
New to This Edition
*Chapters on composite SEM, also called partial least squares path modeling or variance-based SEM; conducting SEM analyses in small samples; and recent developments in mediation analysis.
*Coverage of new reporting standards for SEM analyses; piecewise SEM, also called confirmatory path analysis; comparing alternative models fitted to the same data; and issues in multiple-group SEM.
*Extended tutorials on techniques for dealing with missing data in SEM and instrumental variable methods to deal with confounding of target causal effects.
 
Pedagogical Features
*New self-test of knowledge about background topics (significance testing, regression, and psychometrics) with scoring key and online primers.
*End-of-chapter suggestions for further reading and exercises with answers.
*Troublesome examples from real data, with guidance for handling typical problems in analyses.
*Topic boxes on special issues and boxed rules to remember.
*Website promoting a learn-by-doing approach, including data, extensively annotated syntax, and output files for all the book’s detailed examples.
  "In this ambitious work, Kline thoughtfully and patiently presents diverse perspectives, effectively enlarging the world of SEM while maintaining coherence. The fifth edition's breadth and timeliness make it an easy choice as the primary text in a graduate course on SEM, with readability that journal articles often lack. Researchers will appreciate the book as an entry point to a range of literatures within the SEM world. Kline’s embrace of open-source R software for SEM is very welcome, as it makes the book’s computer examples immediately accessible to readers everywhere."–Edward E. Rigdon, PhD, Marketing RoundTable Professor, Robinson College of Business, Georgia State University
"Kline’s fifth edition is thoroughly updated and greatly expanded. I love the emphasis on Open Science, and I am impressed by the variety of new methodological techniques in SEM that Kline has managed to effectively introduce in the fifth edition. I can’t wait to use this text in my SEM class!"–D. Betsy McCoach, PhD, Neag School of Education, University of Connecticut
"A wonderful introductory book that can be used by individuals without extensive quantitative backgrounds. I use this book to teach an introductory SEM course, but I also use it as a personal reference for my research. It is very readable, which is the number-one reason why I assign this text to my students. I love the walk-through examples with references to real data and syntax. I have always liked Kline’s practical recommendations, and they continue to be really helpful for newbies to SEM–my students constantly reference these sections."–Naomi Ekas, PhD, Department of Psychology, Texas Christian University
"One of the primary strengths of Kline’s book is that it is written in plain English, but with sufficient sophistication that the reader is well prepared to read more technical books or articles on advanced topics. Another strength is the helpful remedies and hints, such as the topic box on the causes of nonpositive definite data matrices and solutions. The most practical advantage of Kline’s text is the exercises at the end of each chapter, and the corresponding answers and explanations."–Stephanie Castro, PhD, College of Business, Florida Atlantic University
"The substantially revised fifth edition lives up to the reputation of prior editions and will be a valuable resource to anyone learning SEM. The online primers are very thorough and give students great refreshers on background topics, including exercises with answers. This edition has appropriate balance between the three 'families' of SEM; I appreciate the detailed descriptions of Pearl’s structural causal model."–Jam Khojasteh, PhD, College of Education and Human Services, Oklahoma State University Rex B. Kline, PhD, is Professor of Psychology at Concordia University in Montréal, Québec, Canada. Since earning a doctorate in clinical psychology, he has conducted research on the psychometric evaluation of cognitive abilities, behavioral and scholastic assessment of children, structural equation modeling, training of researchers, statistics reform in the behavioral sciences, and usability engineering in computer science. Dr. Kline has published a number of chapters, journal articles, and books in these areas.
  Graduate students, instructors, and researchers in psychology, education, human development and family studies, management, sociology, social work, nursing, public health, criminal justice, and communication. Serves as a text for graduate-level courses in structural equation modeling, multivariate statistics, advanced quantitative methods, multiple regression, or causal modeling.
  Introduction
– What’s New
– Book Website
– Pedagogical Approach
– Principles > Software
– Symbols and Notation
– Enjoy the Ride
– Plan of the Book
I. Concepts, Standards, and Tools
1. Promise and Problems
– Preparing to Learn SEM
– Definition of SEM
– Basic Data Analyzed in SEM
– Family Matters
– Pedagogy and SEM Families
– Sample Size Requirements
– Big Numbers, Low Quality
– Limits of This Book
– Summary
– Learn More
2. Background Concepts and Self-Test
– Uneven Background Preparation
– Potential Obstacles to Learning about SEM
– Significance Testing
– Measurement and Psychometrics
– Regression Analysis
– Summary
– Self-Test
– Scoring Criteria
3. Steps and Reporting
– Basic Steps
– Optional Steps
– Reporting Standards
– Reporting Example
– Summary
– Learn More
4. Data Preparation
– Forms of Input Data
– Positive Definiteness
– Missing Data
– Classical (Obsolete) Methods for Incomplete Data
– Modern Methods for Incomplete Data
– Other Data Screening Issues
– Summary
– Learn More
– Exercises
– Appendix 4.a. Steps of Multiple Imputation
5. Computer Tools
– Ease of Use, Not Suspension of Judgment
– Human–Computer Interaction
– Tips for SEM Programming
– Ease of Use, Not Suspension of Judgment
– Commercial versus Free Computer Tools
– R Packages for SEM
– Free SEM Software with Graphical User Interfaces
– Commercial SEM Computer Tools
– SEM Resources for Other Computing Environments
– Summary
II. Specification, Estimation, and Testing
6. Nonparametric Causal Models
– Graph Vocabulary and Symbolism
– Contracted Chains and Confounding
– Covariate Selection
– Instrumental Variables
– Conditional Independencies and Other Types of Bias
– Principles for Covariate Selection
– d-Separation and Basis Sets
– Graphical Identification Criteria
– Detailed Example
– Summary
– Learn More
– Exercises
7. Parametric Causal Models
– Model Diagram Symbolism
– Diagrams for Contracted Chains and Assumptions
– Confounding in Parametric Models
– Models with Correlated Causes or Indirect Effects
– Recursive, Nonrecursive, and Partially Recursive Models
– Detailed Example
– Summary
– Learn More
– Exercises
– Appendix 7.a. Advanced Topics in Parametric Models
8. Local Estimation and Piecewise SEM
– Rationale of Local Estimation
– Piecewise SEM
– Detailed Example
– Summary
– Learn More
– Exercises
9. Global Estimation and Mean Structures
– Simultaneous Methods and Error Propagation
– Maximum Likelihood Estimation
– Default ML
– Analyzing Nonnormal Data
– Robust ML
– FIML for Incomplete Data versus Multiple Imputation
– Alternative Estimators for Continuous Outcomes
– Fitting Models to Correlation Matrices
– Healthy Perspective on Estimators and Global Estimation
– Detailed Example
– Introduction to Mean Structures
– Précis of Global Estimation
– Summary
– Learn More
– Exercises
– Appendix 9.a. Types of Information Matrices and Computer Options
– Appendix 9.b. Casewise ML Methods for Data Missing Not at Random
10. Model Testing and Indexing
– Model Testing
– Model Chi-Square
– Scaled Chi-Squares and Robust Standard Errors for Nonnormal Distributions
– Model Fit Indexing
– RMSEA
– CFI
– SRMR
– Thresholds for Approximate Fit Indexes
– Recommended Approach to Fit Evaluation
– Global Fit Statistics for the Detailed Example
– Power and Precision
– Summary
– Learn More
– Exercises
– Appendix 10.a. Significance Testing Based on the RMSEA
11. Comparing Models
– Nested Models
– Building and Trimming
– Empirical versus Theoretical Respecification
– Chi-Square Difference Test
– Modification Indexes and Related Statistics
– Intelligent Automated Search Strategies
– Model Building for the Detailed Example
– Comparing Nonnested Models
– Equivalent Models
– Coping with Equivalent or Nearly Equivalent Models
– Summary
– Learn More
– Exercises
– Appendix 11.a. Other Types of Model Relations and Tests
12. Comparing Groups
– Issues in Multiple-Group SEM
– Detailed Example for a Path Model of Achievement and Delinquency
– Tests for Conditional Indirect Effects over Groups
– Summary
– Learn More
– Exercises
III. Multiple-Indicator Approximation of Concepts
13. Multiple-Indicator Measurement
– Concepts, Indicators, and Proxies
– Reflective Measurement and Effect Indicators
– Causal–Formative Measurement and Causal Indicators
– Composite Measurement and Composite Indicators
– Mixed-Model Measurement
– Considerations in Selecting a Measurement Model
– Cautions on Formative Measurement
– Summary
14. Confirmatory Factor Analysis
– EFA versus CFA
– Suggestions for Selecting Indicators
– Basic CFA Models
– Other Methods for Scaling Factors
– Detailed Example for a Basic CFA Model of Cognitive Abilities
– Respecification of CFA Models
– Estimation Problems
– Equivalent CFA Models
– Special Tests with Equality Constraints
– Models for Multitrait–Multimethod Data
– Second-Order and Bifactor Models with General Factors
– Summary
– Learn More
– Exercises
– Appendix 14.a. Identification Rules for Correlated Errors or Multiple Loadings
15. Structural Regression Models
– Full SR Models
– Two-Step Modeling
– Other Modeling Strategies
– Detailed Example of Two-Step Modeling in a High-Risk Sample
– Partial SR Models with Single Indicators
– Example for a Partial SR Model
– Summary
– Learn More
– Exercises
16. Composite Models
– Modern Composite Analysis in SEM
– Disambiguation of Terms
– Special Computer Tools
– Motivating Example
– Alternative Composite Model
– Partial Least Squares Path Modeling Algorithm
– PLS PM Analysis of the Composite Model
– Henseler–Ogasawara Specification and ML Analysis
– Summary
– Learn More
– Exercises
IV. Advanced Techniques
17. Analyses in Small Samples
– Suggestions for Analyzing Common Factor Models
– Analysis of a Common Factor Model in a Small Sample
– Controlling Measurement Error in Manifest Variable Path Models
– Adjusted Test Statistics for Small Samples
– Bayesian Methods and Regularized SEM
– Summary
– Learn More
– Exercises
18. Categorical Confirmatory Factor Analysis
– Basic Estimation Options for Categorical Data
– Overview of Continuous/Categorical Variable Methodology
– Latent Response Variables and Thresholds
– Polychoric Correlations
– Measurement Model and Diagram
– Methods to Scale Latent Response Variables
– Estimators, Adjusted Test Statistics, and Robust Standard Errors
– Models with Continuous and Ordinal Indicators
– Detailed Example for Items about Self-Rated Depression
– Other Estimation Options for Categorical CFA
– Item Response Theory and CFA
– Summary
– Learn More
– Exercises
19. Nonrecursive Models with Causal Loops
– Causal Loops
– Assumptions of Causal Loops
– Identification Requirements
– Respecification of Nonrecursive Models That Are Not Identified
– Order Condition and Rank Condition
– Detailed Example for a Nonrecursive Partial SR Model
– Blocked-Error R² for Nonrecursive Models
– Summary
– Learn More
– Exercises
– Appendix 19.a. Evaluation of the Rank Condition
20. Enhanced Mediation Analysis
– Mediation Analysis in Cross-Sectional Designs
– Effect Sizes for Indirect Effects
– Cross-Lag Panel Designs for Mediation
– Conditional Process Analysis
– Causal Mediation Analysis Based on Nonparametric Models and Counterfactuals
– Reporting Standards for Mediation Studies
– Summary
– Learn More
– Exercises
21. Latent Growth Curve Models
– Basic Latent Growth Models
– Data Set for Analyzing Basic Growth Models with No Covariates
– Example Analyses of Basic Growth Models
– Example for a Growth Predictor Model with Time-Invariant Covariates
– Practical Suggestions for Latent Growth Modeling
– Extensions of Latent Growth Models
– Summary
– Learn More
– Exercises
– Appendix 21.a. Unequal Measurement Intervals and Options for Defining the Intercept
22. Measurement Invariance
– Levels of Invariance
– Analysis Decisions
– Partial Measurement Invariance
– Detailed Example for a Two-Factor Model of Divergent Thinking
– Practical Suggestions for Measurement Invariance Testing
– Measurement Invariance Testing in Categorical CFA
– Other Statistical Approaches to Estimating Measurement Invariance
– Summary
– Exercises
23. Best Practices in SEM
– Resources
– Bottom Lines and Statistical Beauty
– Mightily Distinguish Your Work (Be a Hero)
– Family Relations
– Specification
– Identification
– Measures
– Sample and Data
– Estimation
– Respecification
– Tabulation
– Interpretation
– Summary
– Learn More
Suggested Answers to Exercises
References
Author Index
Subject Index
About the Author
 

Additional information