statespace-toolkit 1.2.0

Two new stochastic volatility examples, renumbered examples, and a guide to what to cite. What the library computes is unchanged, and so are the archived packages.

Examples

• ex07: the auxiliary mixture sampler for stochastic volatility, on generated data. It compares the seven-component mixture with the exact density, checks the collapsed Gibbs sampler against the truth, and reweights the draws to the exact posterior.
• ex08: stochastic volatility with MA(1) and Student-t errors, the models of Chan and Hsiao (2014) with mean zero, on daily silver returns.
• Renumbered: SV in mean is now ex09 (was ex07), and TVP-MIDAS is ex10 (was ex08).
• Shorter headers that cite the source of each method.

Documentation

• CITING.md maps every library function, example and archived package to the paper to cite, with BibTeX.
• The README states the tested MATLAB release (R2025b).
• provenance.md gives the published pages of Chan and Hsiao (2014), 155-176.

Full changelog: v1.1.0...v1.2.0

statespace-toolkit 1.1.0

Three new stochastic volatility samplers, a restructured ex05 and a simpler ex07. The archived packages are unchanged.

Library

• ssm.ksc_rw_diffuse: random-walk log-volatility with a mean-zero prior on its first value (twin of bvar.sv.ksc_rw_diffuse).
• ssm.ksc_ar1_mean: stationary AR(1) log-volatility with a mean (twin of bvar.sv.ksc_ar1_mean).
• ssm.ksc_rw_noncentered: the noncentered random-walk log-volatility of Chan (2018).

Each is tested against the published code it comes from.

Examples

• ex05 runs on generated data only; its monthly GDP application has moved to a bvar-toolkit tutorial.
• ex07 estimates a simpler version of the model of Chan (2017).
• Clearer headers and comments throughout.

Documentation: Zenodo DOI in the README and CITATION.cff; corrected license notes.

Full changelog: v1.0.0...v1.1.0

statespace-toolkit 1.0.0

First archived release.

• core/: the ssm MATLAB library for Bayesian state space models: the precision sampler, the integrated likelihood, missing data, and an accept-reject Metropolis-Hastings step, with unit tests.
• replications/: fourteen replication packages, each archived file byte for byte as published.
• examples/: eight short scripts, each running in under a minute.

Requires MATLAB with the Statistics and Machine Learning Toolbox.