Analyzing and Comparing Univariate, Multivariate, and Bifactor Generalizability Theory Designs for Hierarchically...

Area of Science:

• Psychometrics
• Personality Psychology
• Quantitative Psychology

Background:

• Generalizability theory (G theory) is crucial for assessing score reliability.
• Structural equation modeling (SEM) offers flexible frameworks for complex G theory designs.
• The Big Five Inventory (BFI-2) provides a robust measure of personality, including multifaceted subdomains.

Purpose of the Study:

• To demonstrate SEM applications for G theory-based univariate, multivariate, and bifactor models.
• To analyze generalizability of multi-occasion BFI-2 data using these SEM designs.
• To evaluate the utility of different G theory models for understanding personality score reliability.

Main Methods:

• Utilized structural equation models (SEM) to implement G theory designs.
• Analyzed multi-occasion data from the expanded Big Five Inventory (BFI-2).
• Employed univariate, multivariate, and bifactor modeling approaches within SEM.

Main Results:

• SEM effectively represents G theory designs for personality assessment.
• Multivariate and bifactor models provide improved generalizability indices for composite scores.
• Bifactor models enable detailed partitioning of score variance and evaluation of subscale contributions.

Conclusions:

• Accounting for item and occasion effects is vital for accurate generalizability.
• Multivariate and bifactor SEM designs offer advanced insights into personality score reliability and dimensionality.
• The study provides practical guidelines and R code for implementing these advanced G theory analyses.