LoginSign Up
← Back to all sparks
B

bsvarSIGNs

INFRA · APIS
Velocity0.0

Bayesian SVARs with Sign, Zero, and Narrative Restrictions

github.com/bsvars/bsvarsigns ↗

Sign, zero and narrative restrictions brought into the bsvars ecosystem.

bayesian-statisticseconometricsstructural-varmacroeconomicsrcppr-package
◆Current state
bsvarSIGNs estimates structural vector autoregressions identified by sign, zero and narrative restrictions, with the sampler in C++ and the objects, workflows and code structure deliberately matched to the bsvars package. Since the 1.0 launch in mid-2024 the releases have been consolidation: a fix pass, then a vignette, citation metadata and C++ changes to stay ahead of an upcoming compiler check.
◆Where it's heading
The package launched with a published roadmap and the stated intention of intensive development, then spent its next two releases on documentation and compliance rather than new identification schemes. The 2.0 version number is not matched by the changes described under it. What the feed shows is a methods package settling in after launch, not one expanding.
◆Prediction
The roadmap referenced at launch is the only stated plan, and the entries since do not say which part of it is next.

◆Recent moves

  1. 1y ago

    First vignette, citation metadata, C++ check fixes

    The package gets its first vignette and better citation information, plus C++ changes to head off a future check failure. For a methods package the vignette is the substantive part; the 2.0 version number overstates what the note describes.

    View source ↗
  2. 1y ago

    bsvarSIGNs 1.0.1

    A one-line patch pointing at an external issue list, with nothing described in the note itself. It arrives six months after the launch release and three weeks before the next.

    View source ↗
  3. 2y ago

    Launch: sign, zero and narrative restrictions for bsvars

    ⚡ SPARK

    The initial release ships a C++ sampler for SVARs identified by sign, zero and narrative restrictions, with estimation, forecasting and structural analysis, replication data from the source papers, and full object-level integration with bsvars. It extends an existing Bayesian VAR framework into an identification family it did not previously cover.

    View source ↗