Skip to content

FAQ.md

Latest commit

 

History

History
226 lines (174 loc) · 6.65 KB

FAQ.md

File metadata and controls

226 lines (174 loc) · 6.65 KB
Raw

Frequently Asked Questions (FAQ)

General Questions

What is GEPPML?

GEPPML (General Equilibrium PPML) is a methodology for analyzing trade policy using general equilibrium models estimated with the Poisson Pseudo-Maximum Likelihood (PPML) estimator. It was developed by Anderson, Larch, and Yotov (2018).

Why use R instead of Stata?

  • Accessibility: R is free and open-source
  • Flexibility: Easier to customize and extend
  • Integration: Better integration with modern data science workflows
  • Performance: The fixest package is very fast
  • Reproducibility: Easier to share and replicate

Does the R version produce the same results as Stata?

Yes! The R implementation is designed to replicate the Stata results exactly (within numerical precision). You can verify this using the scripts/validation_report.R script.

Installation and Setup

What version of R do I need?

R version 4.0 or higher is recommended. The package may work with older versions but has not been tested.

The script fails with "package not found" error

Install the required packages:

install.packages(c("data.table", "fixest", "haven", "dplyr"))

How do I set the working directory?

If using RStudio:

setwd(dirname(rstudioapi::getActiveDocumentContext()$path))

If not using RStudio:

setwd("C:/path/to/GEPPML_R")  # Replace with your actual path

Running the Analysis

How long does the analysis take?

On a modern computer, the full analysis typically takes:

  • Baseline scenario: < 1 second
  • Conditional scenario: < 1 second
  • Full endowment scenario: 10-30 seconds (depending on convergence)

The convergence is very slow

Try:

  1. Increasing the tolerance: tolerance <- 0.01 (less precise but faster)
  2. Checking your data for errors
  3. Using a better starting point

What if convergence is not achieved?

The script will stop after 100 iterations by default. You can:

  1. Increase max_iterations (e.g., to 200)
  2. Check if results are close enough for your purposes
  3. Investigate the data for potential issues

Can I use my own data?

Yes! Your data should have:

  • exporter: Exporting country code
  • importer: Importing country code
  • trade: Bilateral trade flows
  • Gravity variables (distance, contiguity, etc.)
  • Same structure as the original data

Interpretation

What is the difference between Conditional and Full GE?

  • Conditional GE: Only multilateral resistances adjust; output and expenditure are fixed
  • Full GE: Everything adjusts including output, expenditure, and prices (complete equilibrium)

What does a positive welfare effect mean?

A positive rGDP_full_ch means the country's real GDP increases under the counterfactual scenario (e.g., border removal).

Why are some welfare effects negative?

Some countries may lose from policy changes due to:

  • Trade diversion
  • Changes in multilateral resistance
  • General equilibrium effects

What is sigma and why is it set to 7?

Sigma is the trade elasticity. The value of 7 is commonly used in the literature based on empirical estimates. You can change it based on your sector or research needs.

Customization

How do I analyze a different trade policy?

Modify the counterfactual trade costs in the script. Examples:

Regional Trade Agreement:

dt[, t_ij_ctrf := exp(DIST_est * LN_DIST + CNTG_est * CNTG + 
                      BRDR_est * BRDR - 0.5 * RTA)]

50% reduction in distance friction:

dt[, t_ij_ctrf := exp(DIST_est * LN_DIST * 0.5 + CNTG_est * CNTG + 
                      BRDR_est * BRDR)]

How do I change the reference country?

Replace "DEU" (Germany) with your desired country code:

dt[exporter == "USA", exporter := "ZZZ"]
dt[importer == "USA", importer := "ZZZ"]

Can I add more gravity variables?

Yes! Modify the baseline model specification:

baseline_model <- feglm(
  trade ~ LN_DIST + CNTG + BRDR + YOUR_VARIABLE | exporter + importer,
  data = dt,
  family = poisson()
)

And update the trade costs accordingly.

How do I analyze multiple sectors?

You would need to:

  1. Have sector-specific data
  2. Run the analysis separately for each sector
  3. Aggregate results as needed

This is not currently automated but could be added.

Output and Results

Where are the results saved?

All results are saved in the output/ folder:

  • all_indexes_geppml.csv: Main results (CSV)
  • all_indexes_geppml.rds: Main results (R format)
  • full_static_all.rds: Complete dataset with all variables
  • geppml.log: Execution log

How do I export results to Excel?

Use the example script:

source("scripts/example_analysis.R")

Or install openxlsx:

install.packages("openxlsx")
library(openxlsx)
write.xlsx(results, "output/results.xlsx")

Can I create visualizations?

Yes! See scripts/example_analysis.R for examples using ggplot2.

Troubleshooting

Error: "object 'exporter' not found"

Make sure your data has the required column names exactly as specified.

Error: "contrasts can be applied only to factors"

This usually means a variable that should be numeric is character. Check your data types.

Results seem unrealistic

Check:

  1. Data quality and units
  2. Trade cost specification
  3. Convergence (was it achieved?)
  4. Parameter values (especially sigma)

Memory error with large datasets

Increase R's memory limit:

memory.limit(size = 16000)  # 16GB (Windows only)

Or use a computer with more RAM.

Academic and Citation

How do I cite this package?

See the README for the full citation. Always cite both:

  1. The original Anderson, Larch, and Yotov (2018) paper
  2. This R implementation

Can I use this for my thesis/paper?

Yes! The package is open-source under the MIT License. Please cite appropriately.

Is this peer-reviewed?

The methodology is published in a peer-reviewed journal (The World Economy). This R implementation is a replication package.

Support

Where can I get help?

  1. Check this FAQ
  2. Read the README.md
  3. Look at the example scripts
  4. Open an issue on GitHub
  5. Email: jamiubadmus001@gmail.com

How do I report a bug?

Open an issue on GitHub with:

  • Description of the bug
  • Steps to reproduce
  • Error messages
  • Your R version and OS

Can I request a feature?

Yes! Open an issue with the "enhancement" tag and describe what you need.

Contributing

How can I contribute?

See CONTRIBUTING.md for detailed guidelines. We welcome:

  • Bug reports and fixes
  • Documentation improvements
  • New features
  • Examples and use cases
  • Testing on different platforms

Last Updated: January 19, 2026

Didn't find your answer?