1st Edition

Composite-Based Structural Equation Modeling Analyzing Latent and Emergent Variables

By Jorg Henseler Copyright 2021

    This book presents powerful tools for integrating interrelated composites--such as capabilities, policies, treatments, indices, and systems--into structural equation modeling (SEM). Jörg Henseler introduces the types of research questions that can be addressed with composite-based SEM and explores the differences between composite- and factor-based SEM, variance- and covariance-based SEM, and emergent and latent variables. Using rich illustrations and walked-through data sets, the book covers how to specify, identify, estimate, and assess composite models using partial least squares path modeling, maximum likelihood, and other estimators, as well as how to interpret findings and report the results. Advanced topics include confirmatory composite analysis, mediation analysis, second-order constructs, interaction effects, and importance–performance analysis. Most chapters conclude with software tutorials for ADANCO and the R package cSEM. The companion website includes data files and syntax for the book's examples, along with presentation slides.

    1. Introduction
    1.1. The Nature of Structural Equation Modeling
    1.2. What is Composite-Based SEM?
    1.3. For Which Purpose Should One Use Composite-Based SEM?
    1.3.1. Using Composite-Based SEM for Confirmatory Research
    1.3.2. Using Composite-Based SEM for Explanatory Research
    1.3.3. Using Composite-Based SEM for Exploratory Research
    1.3.4. Using Composite-Based SEM for Descriptive Research
    1.3.5. Using Composite-Based SEM for Predictive Research
    1.3.6. When to Use Composite-Based SEM?
    1.4. Software Tutorial: Getting Started
    1.4.1. First Steps in ADANCO
    1.4.2. First Steps in cSEM
    2. Auxiliary Theories, with Florian Schuberth
    2.1. The Need for Auxiliary Theories
    2.2. Different Types of Science
    2.3. TheAuxiliaryTheory of Behavioral Science: MeasurementTheory
    2.4. The Auxiliary Theory of Design Science: Synthesis Theory
    3. Model Specification
    3.1. What Is A Structural Equation Model?
    3.2. The Outer Model
    3.2.1. Composite Models
    3.2.2. Reflective Measurement Models
    3.2.3. Causal-Formative Measurement Models
    3.2.4. Single-Indicator Measurement Models
    3.2.5. Categorical Variables
    3.3. The Inner Model
    3.4. Software Tutorial: Model Specification
    3.4.1. Specifying Structural Equation Models in ADANCO
    3.4.2. Specifying Structural Equation Models in cSEM
    4. Model Identification
    4.1. The Necessity of Identification
    4.2. Ensuring Model Identification in Composite-Based SEM
    4.3. Ensuring Empirical Identification in Composite-Based SEM
    4.4 ‘Chance Correlations’
    4.4.1. The Problem with ‘Chance Correlations’
    4.4.2. Avoiding ‘Chance Correlations’
    4.5. The Dominant Indicator Approach As a Solution to Sign Indeterminacy
    4.6. Identification Rules
    5. Model Estimation
    5.1. Composite-Based Estimators for Composite Models
    5.1.1. Stand-Alone Constructions: Sum Scores, Preset Weights, and Principal Components
    5.1.2. The Partial Least Squares Path Modeling Algorithm
    5.1.3. Generalized Structured Component Analysis
    5.2. Composite-Based Estimators for Reflective Models
    5.2.1. Consistent Partial Least Squares
    5.2.2. Sum Scores with Correction for Attenuation
    5.3. Fitting Functions
    5.4. Tutorial: Model Estimation
    5.4.1. Estimating Models Using ADANCO
    5.4.2. Estimating Models Using cSEM
    6. Global Model Assessment: Model Fit
    6.1. The Motivation for Model Fit
    6.2. Model Fit Tests
    6.2.1. Non-Parametric Model Fit Tests
    6.2.2. Parametric Model Fit Tests
    6.3. Model Fit Indices
    6.3.1. Standardized Root Mean Squared Residual (SRMR)
    6.3.2. Root Mean Square Residual Covariance (RMSθ)
    6.3.3. Fit Measures Provided by Covariance-based SEM
    6.4. What If Model Fit Is Low?
    6.5. Beware of Alleged Goodness of Fit Indices
    6.5.1. Four “Goodness of Fit Indices” That Are Not Model Fit Indices
    6.5.2. The Different Meanings of Fit
    6.6. Tutorial: Model Testing
    6.6.1. Using ADANCO for Model Testing
    6.6.2. Using cSEM for Model Testing
    7. Local Model Assessment
    7.1. The Need for Reliability and Validity
    7.2. Assessing Composite Models of Emergent Variables
    7.2.1. Nomological Validity
    7.2.2. The Reliability of Composites
    7.2.3. Weights
    7.3. Assessing Reflective Measurement Models of Latent Variables
    7.3.1. Construct Validity
    7.3.2. Unidimensionality
    7.3.3. Discriminant Validity
    7.3.4. Reliability of Construct Scores
    7.4. Assessing Causal-Formative Measurement Models
    7.5. Assessing Inner Models
    7.5.1. R2 and Adjusted R2
    7.5.2. Inter-Construct Correlations
    7.5.3. Path Coefficients
    7.5.4. Indirect Effects
    7.5.5. Total Effects
    7.5.6. Effect Size (Cohen’s f 2)
    7.6. Inferential Statistics and the Bootstrap
    7.7. Construct Scores
    7.8. What If There Is No Output?
    7.9. Tutorial: Model Assessment
    7.9.1. Model Assessment Using ADANCO
    7.9.2. Model Assessment Using cSEM
    8. Confirmatory Composite Analysis, with Florian Schuberth
    8.1. Motivation
    8.2. Confirmatory Composite Analysis: Model Specification
    8.3. Confirmatory Composite Analysis: Model Identification
    8.4. Confirmatory Composite Analysis: Model Estimation
    8.5. Confirmatory Composite Analysis: Model Testing
    8.6. Tutorial: Confirmatory Composite Analysis
    8.6.1. Confirmatory Composite Analysis Using ADANCO
    8.6.2. Confirmatory Composite Analysis Using cSEM
    9. Mediation Analysis
    9.1. The Logic of Mediation
    9.2. Mediation Analysis Using Composite-based SEM
    9.3. Tutorial: Mediation Analysis
    9.3.1. Mediation Analysis Using ADANCO
    9.3.2. Mediation Analysis Using cSEM
    10. Second-Order Constructs
    10.1. A Typology of Second-Order Constructs and Their Use
    10.2. Modeling Type-I Second-Order Constructs: LatentVariablesMeasured by Latent Variables
    10.3. Modeling Type-II Second-Order Constructs: Emergent Variables Made of Latent Variables
    10.4. Modeling Type-III Second-Order Constructs: Latent Variables Measured by Emergent Variables
    10.5. Modeling Type-IV Second-Order Constructs: Emergent Variables Made of Emergent Variables
    10.6. Modeling Type-V Second-Order Constructs: Latent Variables Measured by Different Types of Variables
    10.7. Modeling Type-VI Second-Order Constructs: Emergent Variables Made of Different Types of Variables
    10.8. Tutorial: Second-Order Constructs
    10.8.1. Modeling Second-Order Constructs with ADANCO
    10.8.2. Modeling Second-Order Constructs with cSEM
    11. Analyzing Interaction Effects
    11.1. The Logic of Interaction Effects
    11.2. Estimating Interaction Effects with Composite-Based SEM
    11.2.1. Multigroup Analysis
    11.2.2. The Two-Stage Approach for Analyzing Interaction Effects
    11.2.3. The Orthogonalizing Approach for Analyzing Interaction Effects
    11.3. Visualizing Interaction Effects
    11.3.1. Surface Analysis
    11.3.2. Spotlight Analysis
    11.3.3. Floodlight Analysis
    11.4. Three-way Interactions
    11.5. Nonlinear Effects
    11.6. Tutorial: Interaction Effects
    11.6.1. Analyzing Interaction Effects Using ADANCO
    11.6.2. Analyzing Interaction Effects Using cSEM
    12. Importance–Performance Analysis
    12.1. Nature and Fields of Application
    12.2. A Step-by-Step Guide to Conducting IPA Using Composite-Based SEM
    12.3. Tutorial: Importance–Performance Analysis
    12.3.1. Using ADANCO for Importance–Performance Analysis
    12.3.2. Using cSEM for Importance–Performance Analysis
    Author Index
    Subject Index
    About the Author


    Jörg Henseler, PhD, is Full Professor and Chair of Product–Market Relations in the Faculty of Engineering Technology of the University of Twente in The Netherlands. He is also Visiting Professor at NOVA Information Management School, NOVA University of Lisbon, Portugal, and Distinguished Invited Professor in the Department of Business Administration and Marketing at the University of Seville, Spain. His broad-ranging research interests encompass empirical methods of marketing and design research as well as the management of design, products, services, and brands. A highly cited researcher, Dr. Henseler is a leading expert on partial least squares (PLS) path modeling, a composite-based structural equation modeling (SEM) technique that bridges design and behavioral research. He has written dozens of scholarly articles, edited or authored several books, served as guest editor for three special journal issues, and chaired conferences on PLS. He serves on several journal editorial boards and has been an invited speaker on SEM at universities around the world. Dr. Henseler chairs the scientific advisory board for the ADANCO software program and regularly provides seminars on PLS path modeling at the PLS School.

    "This book offers a novel perspective on SEM in which constructs are viewed as composites rather than factors. Explaining the principles and rationales of composite-based SEM, the book provides examples and tutorials using the software ADANCO and the R package cSEM. It will make a great supplementary text for graduate-level courses on SEM, latent variable modeling, and multivariate analysis. It is written in such a way that both beginning doctoral students and experienced researchers can learn a lot from it. Students will get a fresh view of SEM and learn the SEM fundamentals; experienced researchers can use this book as an opportunity to elevate the discussion about SEM."--Ge Jiang, PhD, Department of Educational Psychology, University of Illinois at Urbana–Champaign

    "Henseler gives an excellent introduction to composite-based SEM, the third member of the SEM family that also includes covariance-based SEM, widely known among researchers in psychology and related fields, and Judea Pearl’s structural causal model, familiar to researchers in epidemiology and medicine. Methods in composite-based SEM offer researchers additional options for modeling, estimating, and testing hypotheses about latent variables, including emergent variables, or proxies for hypothetical constructs analyzed as linear combinations of observed variables. With clear descriptions, numerous examples, and coverage of advanced topics such as moderation, mediation, and importance–performance analysis, Henseler’s book is a great starting point for both students and established researchers seeking to expand their modeling repertoires."--Rex B. Kline, PhD, Department of Psychology, Concordia University, Canada

    "This extraordinary work on SEM to estimate composite models provides up-to-date explanations illustrated with a range of good examples. The book is 'ambidextrous' in the way it is both student-friendly and rigorous. It will surely make an impact on the SEM community. I plan to use this text in my master's- and doctoral-level SEM courses, as well as in my consulting activities with companies.”--Jose Benitez, PhD, Rennes School of Business, France; and School of Business and Economics, University of Granada, Spain-