confirmatory factor analysis

Existing research frequently suffers from model misspecification of formative constructs, and the absence of a consensus-based validation methodology leads scholars to erroneously apply reflective measurement frameworks, thereby compromising construct validity. Method: This paper introduces the first dedicated, multi-stage validation framework for formative constructs, integrating systematic literature review, descriptive statistics, multicollinearity diagnostics, and formative-model-specific tests to rigorously distinguish formative (causal) from reflective (effect) measurement logic. Contribution/Results: The framework ensures both theoretical rigor and practical feasibility, substantially enhancing the psychometric soundness and statistical integrity of formative indicators. It provides a reproducible, defensible methodological pathway for scale development and construct validation, directly addressing longstanding measurement challenges in behavioral and social science research.

This study addresses the common practice in confirmatory factor analysis of accepting standardized factor loadings as low as 0.50, which leads to elevated measurement error, compromised construct validity, and unstable factor solutions. Building on the logic of average variance extracted (AVE) and communality, the authors propose and justify a uniform item-level threshold of λ ≥ 0.70, aligning it with construct-level validity requirements. Through theoretical derivation, Monte Carlo simulations, and structural equation modeling, the research systematically evaluates the impact of weak loadings on measurement quality, factor score determinacy, and model fit. Findings demonstrate that retaining indicators with λ < 0.70 significantly undermines model accuracy and robustness, whereas enforcing the λ ≥ 0.70 criterion enhances the explanatory power and overall quality of latent variable models.

Conventional factor models rely heavily on mean and covariance fit assessment, often overlooking critical assumptions—such as the distributional form of latent variables and the functional form of observed indicators—leading to insensitive model diagnostics. Method: This paper systematically extends generalized residual theory to common factor models with continuous and mixed-type observed variables, developing a novel fit evaluation framework that simultaneously tests assumptions about latent variable distributions and indicator functional forms. It introduces generalized-residual-based diagnostic statistics, validated through Monte Carlo simulations and empirical analysis. Contribution/Results: The proposed method effectively detects structural misspecifications—e.g., nonnormal latent distributions or nonlinear indicator relationships—that conventional goodness-of-fit indices (e.g., χ², CFI, RMSEA) fail to identify. It substantially enhances the sensitivity, interpretability, and diagnostic accuracy of model fit assessment, offering a more rigorous and assumption-robust approach to evaluating factor model validity.

To address the computational complexity and convergence difficulties inherent in joint estimation of measurement and structural models in item response theory (IRT), this paper proposes a two-step maximum likelihood estimation procedure: first, estimating measurement model parameters independently; second, estimating structural model parameters with measurement parameters held fixed. This work provides the first systematic theoretical justification—under settings involving continuous latent variables and categorical observed variables—of the statistical consistency, robustness, and computational efficiency of the two-step approach. Compared to conventional one-step estimation (prone to non-convergence) and three-step methods (susceptible to bias accumulation), the proposed method offers conceptual clarity, implementation simplicity, reliable standard errors, and stable convergence. Extensive simulation studies and empirical analyses validate its efficacy and generalizability across diverse latent variable models. The framework establishes a novel, general-purpose, flexible, and practical estimation paradigm for educational measurement, psychometrics, and related fields.

Existing structural equation modeling (SEM) frameworks struggle to model latent variable variances that depend on other latent variables, thereby limiting the characterization of latent heteroscedasticity—such as in psychological constructs like personality or creativity. To address this, we propose Bayesian Gaussian Distributional SEM, the first SEM extension integrating distributional regression into the SEM framework to jointly model both the mean and variance of latent variables. Leveraging Bayesian inference and MCMC sampling, our approach flexibly specifies latent variances as arbitrary functions of other latent variables. Simulation studies demonstrate high statistical reliability and computational efficiency. Empirical analysis of personality data reveals that emotional stability significantly moderates the variability of neuroticism—a finding inaccessible under conventional SEM. This work introduces a novel theoretical tool and methodological paradigm for modeling latent heteroscedasticity, advancing both substantive theory testing and statistical methodology in behavioral and social sciences.

Existing multidimensional factor models struggle to accommodate the typical 3–5 dimensional latent constructs in psychometrics and lack a unified framework for modeling multiparameter moderation effects. This paper proposes a scalable penalized maximum likelihood estimation method applicable to arbitrarily many factors, enabling— for the first time—the joint estimation of linear and nonlinear moderation effects within high-dimensional models. By incorporating ridge, lasso, and alignment penalties, the approach simultaneously stabilizes parameter estimation, detects partial measurement noninvariance, and enhances interpretability. Leveraging closed-form analytical gradients, the method avoids computationally intensive numerical integration and MCMC sampling, substantially improving computational efficiency. Simulation and empirical studies demonstrate accurate recovery of complex moderation patterns. The proposed method provides a scalable, efficient, and robust new tool for measurement invariance research involving multidimensional constructs.

This study addresses the long-standing divide between latent variable models and network models in psychometrics, which has hindered theoretical integration and methodological innovation. Through an exploratory literature review, cross-disciplinary comparison of statistical models, and visualization techniques, it systematically examines the intrinsic connections among Item Response Theory (IRT), Structural Equation Modeling (SEM), Generalized Linear Models (GLM), and network analysis. The work proposes a unified modeling paradigm that elucidates both the commonalities and complementarities across these approaches, establishing an integrative framework bridging latent variable and network perspectives. This framework not only offers a novel lens for addressing longstanding debates about the nature of psychological constructs but also facilitates the development of reproducible, modular psychometric tools, thereby advancing interdisciplinary collaboration and methodological synthesis.

Existing structural equation modeling (SEM) approaches face multiple limitations in modeling composite variables—such as indices, formative constructs, and bundle variables—including inability to represent their construction process, fixed (non-estimated) weights, difficulty in specifying them as endogenous, and lack of capacity to test full mediation or component-variable influence. This paper introduces two novel methods grounded in the H-O specification, integrating phantom variables with pseudo-indicator techniques. For the first time, these methods treat inverse or full weights as freely estimated parameters, enabling flexible inclusion of composite variables at any model position—including endogenous—and rigorous testing of effect transmission. By unifying linear combination weight estimation with the SEM framework, the methods successfully estimate weights, model composites endogenously, and distinguish mediation from formative mechanisms in empirical data. This significantly enhances modeling accuracy, interpretability, and guidance for model selection in multivariate behavioral research.

This paper addresses the identification challenge in multidimensional continuous measurement error models, where all observed variables are contaminated by latent, mutually dependent errors and no injective mapping is available. Methodologically, we construct a third-order cross-moment tensor and integrate Kruskal decomposition, the Kotlarski identity, and integral operator theory; we introduce the novel concept of “signal rank” and generalize Kruskal rank, achieving global identification of the joint distribution of latent variables and measurement errors for the first time under non-injective, multidimensional nonlinear settings. Our contribution lies in breaking free from conventional reliance on clean measurements or injectivity assumptions, providing empirically testable identification conditions, and substantially broadening the applicability of latent variable models. The proposed framework delivers a robust and general theoretical foundation for modeling noisy data in economics, psychology, and related disciplines.

This study addresses a critical limitation in cross-cultural research: single-item measures lack latent variable proxies and multiple indicators, rendering conventional tests for differential item functioning (DIF) or measurement invariance (MI) infeasible. To overcome this challenge, the authors propose a penalized heteroskedastic ordered probit model that integrates regularization techniques to resolve identification and estimation issues inherent in single-item data. This approach enables, for the first time, a viable framework for conducting DIF and MI analyses with single-item responses, thereby circumventing the traditional reliance on multi-item scales or multiple indicators. The method offers a novel analytical tool for cross-cultural comparisons in resource-constrained settings or contexts where only single-item measures are available, significantly expanding the scope of valid cross-cultural inference under practical constraints.