RSDC: Regime-Switching Dynamic Correlation Models in R

RSDC is an R package for modeling time-varying asset correlations through a regime-switching framework. It supports flexible multivariate correlation dynamics with either fixed or time-varying transition probabilities (TVTP) driven by exogenous covariates.

This is particularly useful in financial applications where asset return correlations exhibit structural breaks or evolve under different market regimes.

Features

• Three model families through one front end, rsdc_estimate(): constant correlations, fixed transition probabilities, and covariate-driven transitions (TVTP) with a logistic link for two regimes and a softmax for more
• Hamilton filter and smoother implemented in C++ (Rcpp/RcppArmadillo), validated against a pure-R reference to ~1e-8 in the test suite
• A global search that operates in a canonical partial-correlation space, in which every candidate is a valid positive-definite correlation matrix, so maximum likelihood stays feasible as the cross-section grows; plus data-driven warm starts (rsdc_starts()) and a multi-start replication diagnostic
• A full "rsdc_fit" S3 interface: print, summary, coef, logLik (so AIC/BIC work), vcov, confint, predict, simulate, plot, broom tidiers and ggplot2::autoplot()
• Standard errors from observed information, OPG, sandwich (QML) or a parametric bootstrap; uncertainty bands on the correlation path
• Forecasting and decoding: rsdc_forecast() (in-sample or strictly out-of-sample), rsdc_forecast_ahead() (multi-step), rsdc_viterbi()
• Portfolio construction from the implied covariances: rsdc_minvar(), rsdc_maxdiv()
• Shipped data: greenbrown, mccc, ff5ind

Installation

install.packages("RSDC")

Development version:

devtools::install_github("ArdiaD/RSDC")

Please cite the package in publications!

By using RSDC, you agree to the following:

1. Include a footnote or reference to the GitHub page:
   https://github.com/ArdiaD/RSDC
2. You assume all risk for using this software.

References

Engle, R.F. (2002). Dynamic conditional correlation: A simple class of multivariate generalized autoregressive conditional heteroskedasticity models. Journal of Business & Economic Statistics, 20(3), 339–350.

Pelletier, D. (2006). Regime switching for dynamic correlations. Journal of Econometrics, 131(1–2), 445–473.

Hamilton, J.D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57(2), 357–384.

Companion article

A companion manuscript (in preparation for The R Journal) presents the methodology, the estimation strategy, a Monte Carlo validation, and a strictly out-of-sample study of five Fama–French industry portfolios with regime-dependent minimum-variance and maximum-diversification allocations. It is developed separately and is fully reproducible from data shipped with this package. Use citation("RSDC") for the current reference.

For a short tour of the API, see the package vignette: browseVignettes("RSDC").