Plan: Dynamic Panel BMA for Economic Growth (v2)

Confirmed: 2026-04-01 Language: R Theme: Dark navy

Topic

Dynamic panel Bayesian Model Averaging: identifying growth determinants in 73 countries across 4 decades (1970–2000) using the bdsm R package. Analysis question: “Which of 9 candidate growth determinants truly drive economic growth, accounting for reverse causality?”

Framing

Descriptive/variable selection analysis (not causal estimation). BMA identifies robust determinants via posterior inclusion probabilities but does not estimate causal effects.

Data Preparation Improvement

3-step pipeline (vs v1’s 2-step):

1. Standardize regressors (scale)
2. Demean by year (time FE removal)
3. Demean by country (entity FE removal)

Script Sections

• 0. Setup (packages, seed, colors, dark theme)
• 1. Data loading + CSV export
• 2. Data preparation (3-step) + CSV export
• 3. Model space (precomputed 2^9 = 512 models)
• 4. Benchmark: kitchen-sink FE regression + CSV export
• 5. BMA default prior (EMS = 4.5) + CSV export
• 6. Built-in bdsm visualizations (3 PNGs)
• 7. Best models (top 8) + CSV export
• 8. Prior sensitivity (EMS 2, 4.5, 8; dilution) + CSV export
• 9. Jointness analysis (HCGHM, LS, DW) + CSV export
• 10. Custom ggplot2 dark-theme figures (3 PNGs)
• 11. Summary

Expected Outputs

• 7 PNG figures
• 10 CSV tables
• README.md
• plan.md (this file)