Testing informative hypotheses in factor analysis models using bayes factors.

Area of Science:

• Psychometrics
• Statistical Modeling
• Bayesian Inference

Background:

• Confirmatory Factor Analysis (CFA) models are widely used in psychology and social sciences.
• Testing specific, theory-driven hypotheses within CFA models presents analytical challenges.
• Existing methods may not adequately capture nuanced theoretical expectations regarding model parameters.

Purpose of the Study:

• To propose a novel Bayesian approach for testing informative hypotheses in CFA.
• To introduce an adjusted fractional Bayes factor for quantifying evidence supporting these hypotheses.
• To demonstrate the application and interpretation of this method using simulation studies and a real-world example.

Main Methods:

• Formulation of informative hypotheses using constrained loadings in CFA.
• Specification of prior distributions using a portion of the data.
• Computation of the adjusted fractional Bayes factor using Markov chain Monte Carlo (MCMC) methods.

Main Results:

• The proposed Bayesian approach effectively quantifies support for informative hypotheses in CFA.
• Simulation studies demonstrate the performance and utility of the adjusted fractional Bayes factor.
• The method allows for direct testing of theoretical expectations regarding reliability, validity, and indicator importance.

Conclusions:

• The Bayesian framework offers a powerful tool for hypothesis testing in CFA.
• The adjusted fractional Bayes factor provides a robust measure of evidence for theory-driven models.
• This approach enhances the ability of researchers to formally evaluate specific theoretical predictions.