Florian Huber

Professor of Economics
University of Salzburg

Department of Economics
Mönchsberg 2a
5020 Salzburg

This page collects software packages and replication code for my research, written in R and Rcpp. The replication packages are self-contained and reproduce the main results of the corresponding paper on simulated data. Note that these models are fairly complex and some knowledge of Bayesian econometrics / statistics is needed to understand the effect of priors. I do not offer any support for these codes. Although I try to keep them error-free, I can’t guarantee this — if you find an error, please contact me.

R packages

  • BGVAR — R toolbox for fast and easy estimation of Bayesian GVAR models. Includes several priors, stochastic volatility, and functions for forecasting and structural analysis. CRAN (with a detailed vignette).

Replication code

  • BART-based VARs — R code for estimating various BART-based VAR versions (BART, mixBART, flexBART, fullBART) with factor stochastic volatility, as proposed in Clark, Huber, Koop, Marcellino and Pfarrhofer (2023, International Economic Review). GitHub.

  • Factor-augmented BART VARs, with T. Clark and G. Koop (Journal of Business & Economic Statistics, forthcoming; DOI) — R code for a flexible Bayesian VAR augmented with nonlinear BART-based factors and stochastic volatility, featuring sign-identified structural analysis. Ships the factorBART R package plus precomputed caches so all figures regenerate in seconds. GitHub.

  • Bayesian Nonparametric VARs — R code for Bayesian VARs with Dirichlet-process mixture shocks and optional idiosyncratic stochastic volatility, as in Huber and Koop (2024, Journal of Applied Econometrics). GitHub.

  • Forecasting US Inflation using Bayesian Nonparametric Models — R code for inflation forecasting with GP-subspace, BART, and UCSV specifications under Dirichlet-process mixture shocks, as in Clark, Huber, Koop and Marcellino (2024, Annals of Applied Statistics). GitHub.

  • Gaussian Process VARs, with N. Hauzenberger, M. Marcellino and N. Petz (JBES, forthcoming) — R code for a nonparametric VAR with Gaussian-process conditional means, equation-wise stochastic volatility, and horseshoe-shrunk structural covariance. GitHub.

  • TVP-BART VARs, with N. Hauzenberger, G. Koop and J. Mitchell (Annals of Applied Statistics, forthcoming) — R code for a semiparametric TVP-VAR in which time-varying coefficients and error covariances are driven by Bayesian Additive Regression Trees (BART). GitHub.

  • Nowcasting in a Pandemic using Non-Parametric Mixed-frequency VARs, with G. Koop, M. Pfarrhofer, L. Onorante and J. Schreiner (Journal of Econometrics, forthcoming). Estimates the mixed-frequency BAVART model and more general BART-based VARs via an SVD-based algorithm with mixture state-equations. GitHub.

  • Investigating Growth-at-Risk Using a Multicountry Non-parametric Quantile Factor Model, with T. Clark, G. Koop, M. Marcellino and M. Pfarrhofer (JBES, 2024). R code implementing the QF-BART model for jointly estimating quantiles of GDP growth across multiple countries. GitHub.

  • Adaptive Shrinkage in Bayesian Vector Autoregressive Models, with M. Feldkircher (JBES, 2019). Estimates a VAR with a hierarchical Normal-Gamma shrinkage prior (and alternative priors) on the autoregressive coefficients with stochastic volatility. GitHub.

  • Approximate Bayesian Inference and Forecasting in Huge-dimensional Multi-country VARs, with M. Feldkircher, G. Koop and M. Pfarrhofer. Implements the Integrated Rotated Gaussian Approximation (IRGA) strategy for very large multi-country VARs combining Horseshoe-prior MCMC with VAMP. GitHub.

  • Combining Shrinkage and Sparsity in Conjugate Vector Autoregressive Models, with N. Hauzenberger and L. Onorante (Journal of Applied Econometrics, 2020). Conjugate Bayesian VAR with Minnesota dummy-observation priors and SAVS post-processing sparsification. GitHub.

  • Subspace Shrinkage in Conjugate Bayesian Vector Autoregressions, with G. Koop (Journal of Applied Econometrics, forthcoming). Conjugate matrix-Normal / Inverse-Wishart VAR with a convex combination of a Minnesota prior and a principal-component subspace projection. GitHub.

  • Threshold Cointegration in International Exchange Rates: A Bayesian Approach, with T. O. Zörner (International Journal of Forecasting, 2019). Estimates a three-regime threshold Bayesian vector error correction model (TBVECM). GitHub.

  • Fast and Flexible Bayesian Inference in Time-Varying Parameter Regression Models, with N. Hauzenberger, G. Koop and L. Onorante (JBES, 2021). SVD-based fast sampler for TVP regressions with sparse-mixture g-prior on the coefficients. GitHub.

  • Should I Stay or Should I Go? A Latent Threshold Approach to Large-scale Mixture Innovation Models, with G. Kastner and M. Feldkircher (Journal of Applied Econometrics, 2019). Latent-threshold TVP-VAR with stochastic volatility; includes the threshtvp R package source. GitHub.

  • Inducing Sparsity and Shrinkage in Time-Varying Parameter Models, with G. Koop and L. Onorante (JBES, 2021). TVP-VAR with global-local shrinkage priors (Horseshoe, LASSO, Normal-Gamma, SSVS, Dirichlet-Laplace) and SAVS post-processing for exact sparsification of time-varying coefficients. GitHub.