Bayesian latent variable models for the analysis of experimental psychology data

Area of Science:

• Psychology
• Statistics
• Data Analysis

Background:

• Experimental psychology often generates complex multivariate data.
• Standard analytical methods may not fully capture the nuances of such data.

Purpose of the Study:

• To explore the application of Bayesian factor analysis and structural equation models for experimental psychology.
• To provide a unified framework for analyzing multivariate data from psychological experiments.
• To compare these advanced models with simpler methods.

Main Methods:

• Review of Bayesian factor analysis and structural equation models.
• Discussion of parameter identification issues.
• Detailed explanation of model estimation using JAGS.
• Bayes factor estimation techniques.

Main Results:

• Demonstrated the utility of Bayesian models for multivariate psychological data.
• Successfully re-analyzed experimental data on risky choice.
• Highlighted the advantages over simpler analytical approaches.

Conclusions:

• Bayesian factor analysis and structural equation models are valuable, albeit non-standard, tools for psychological research.
• These models facilitate unified and robust inference from complex experimental data.
• The approach offers a powerful alternative for analyzing data from multiple experimental stimuli.