Hello R Community,
I'm excited to announce the first stable release of convergenceDFM, an R package now available on CRAN. This package is designed for researchers and analysts working with high-dimensional Dynamic Factor Models (DFMs).
What does it do?
Drawing from the wiki documentation, convergenceDFM provides a comprehensive suite of tools to diagnose and assess convergence in DFM estimations. The core functions help you:
- Calculate and visualize key convergence diagnostics.
- Monitor parameter stability across iterations.
- Implement various heuristics to determine if a model has reliably converged.
At its core, the package addresses situations where you need to determine if disaggregated economic variables (e.g., sectoral prices, regional outputs) exhibit mean-reversion toward each other—a question central to economic theory but notoriously difficult to answer with standard tools. The solution combines dimension reduction, state-space modeling, and Bayesian inference to deliver rigorous convergence tests with full uncertainty quantification.
Specifically, this library uses Factor Ornstein-Uhlenbeck linear processes to quantify mean-reversion speeds and cross-factor coupling via Bayesian MCMC.
Why is it useful?
Ensuring convergence is a critical but often overlooked step in factor modeling. This package aims to make the process more transparent, reproducible, and accessible, moving beyond simple eyeball checks of trace plots.
Getting Started:
- Install from CRAN:
install.packages("convergenceDFM") - Browse the Code & Wiki: GitHub - IsadoreNabi/convergenceDFM
The package vignettes and wiki provide detailed examples, from basic usage to advanced scenarios. I hope convergenceDFM becomes a valuable tool in your macroeconometric or high-dimensional time series workflow.
I welcome any feedback, bug reports, or contributions on GitHub. What are your experiences or challenges with convergence diagnostics?
Best,
José Mauricio Gómez Julián