Standardized Generalized Hierarchical Factor Models (GenHFM)

This repository contains all the data, R code, Stan models, and supplementary materials associated with the paper: "A Unified Framework for Psychometrics in Experimental Psychology: The Standardized Generalized Hierarchical Factor Model".

Note: Due to the large size of certain output files (fitted models and large .RData objects), the heaviest results are hosted on our OSF project [https://osf.io/5dutv/].

---

📂 Repository Structure

The directory is organized as follows:

1. Data

Contains the datasets used across the analyses.

• 1.1 Processed: Cleaned data ready for model fitting.
• 1.2 Raw data: Original raw datasets.

2. Figures

Contains all figures presented in the main text of the manuscript.

3. Manuscript

Drafts of the manuscript and Supplementary Material documents.

4. Model diagrams

Images and schematics illustrating the GenHFM architecture.

5. R scripts

The complete R codebase for reproducing the entire study.

• 5.1 Appendix: Code for generating the unit vector visualization.
• 5.2 Empirical analysis:
  • 5.2.1 Rey-Mermet: Code to fit GenHFMs using criteria from Rey-Mermet et al. (2018) and posterior predictive check (PPC) figures.
    • datasets: Original data from Rey-Mermet et al.
    • DDM_fitted_models: Pre-fitted DDM models (note: these may be hosted on OSF due to size constraints).
  • 5.2.2 Viviani: Scripts for data cleaning, model fitting, and result analysis.
  • 5.2.3 Whitehead: Scripts for data cleaning, model fitting, and result analysis.
• 5.3 Meta-analysis: Code for fitting ANOVA metamodels and estimating parameter recovery indices.
• 5.4 Model Diagrams: R code used to generate the model schematics and export individual images.
• 5.5 R functions: Custom functions used throughout the empirical analyses, meta-analysis, and simulations.
• 5.6 Simulation study: Comprehensive code for running the simulation, generating plots, and analyzing recovery.
• 5.7 Supplementary Material: Code used to generate results specifically for the supplement.

6. Results

Summary outputs and saved R objects.

• 6.1 Rdata:
  • 6.1.1 ELPDs: Results for model comparison using Expected Log-Pointwise Predictive Density.
  • 6.1.2 LMM: Fitted Linear Mixed Models for the Viviani dataset.
  • 6.1.3 Meta-parameters: Results and estimates from the meta-analysis/ANOVA.
  • 6.1.4 Simulation study: Raw and processed results from the simulation study.
  • 6.1.5 Supplementary Material: Result objects for the supplementary material.
  • 6.1.6 Tables: Results tables included in the main manuscript.

7. Stan models

The source code for the Bayesian models.

• 7.1 BGHFM: The primary Generalized Hierarchical Factor Model (GenHFM) implementations.
• 7.2 BGHFM (sim study): Versions of the GenHFM optimized specifically for the simulation study.
• 7.3 BLMM: Bayesian Linear Mixed Models used in the meta-analysis.

---

🚀 How to Reproduce

To run the scripts in this repository, you will need R, Stan, and several R packages (primarily cmdstanr, posterior and tidyverse).

To ensure all scripts run correctly without the need to manually set working directories, simply:

1. Download or clone the entire repository.
2. Open the GenHFM.Rproj file in RStudio.
3. Run the desired scripts; all file paths are relative to the project root.

🔗 External Links

• OSF Repository: [https://osf.io/5dutv/] (Contains large .rds model objects and high-volume datasets).

---