BVAR toolkit for macroeconomic forecasting¶

A versatile package for Bayesian Vector Autoregressions (BVARs). It supports macroeconomic forecasting with Bayesian shrinkage, marginal likelihood, cross-validation, and conditional forecasting.

Features¶

• Natural conjugate (Normal-Inverse-Wishart) setting with Minnesota shrinkage.
• Sum-of-coefficients and single-unit-root priors via dummy observations.
• Hyperparameter optimisation — marginal likelihood (GLP, 2015) or cross-validation.
• COVID-19 dummies following Cascaldi-Garcia (2022).
• Conditional & unconditional forecasting with hard, soft, and skewed constraints.
• Generalised Impulse Response Functions (Pesaran & Shin, 1998).
• Forecast revision analysis for counterfactual comparisons.
• Nowcasting uncertainty — treats nowcasts as soft constraints.

Quick Start¶

See First forecast in ten lines in the user guide.

Installation¶

pip install bvar

Project Layout¶

src/bvar/           # Source code
docs/               # Documentation, notebooks & this site
tests/              # Unit and integration tests

References¶