BVARSVResult#

class cultivars.multivariate.large_dim.volatility.BVARSVResult(*, endog, names, order, trend, prior_label, coefficients, deterministic, beta_mean, sigma_u, beta_draws, sigma_draws, h_draws, impact_draws, vol_of_vol, resid, fittedvalues, nobs, n_draws, n_burn, thin)[source]

Bases: _VectorPosteriorDrawsResult

A fitted stochastic-volatility BVAR: one system, a path of covariances.

The family’s shared propagation surface, with the covariance read at the end of the sample: sigma_u and the covariance draws are Sigma_T = A^{-1} diag(exp h_T) A^{-T}, the covariance relevant for what comes next, and the predictive propagates each draw’s log-variance random walk forward from there. The full history lives in h_draws and volatility_path().

The triangular factorization is taken in the order of names at estimation time, so – as for the TVP-SV model – the recursive ordering is a structural declaration already made; orthogonalized responses inherit it.

Variables:

  • prior_label – Short description of the prior estimated under.

  • h_draws – (S, nobs, k) kept log-variance path draws.

  • impact_draws – (S, k, k) kept draws of the unit-lower A^{-1}.

  • vol_of_vol – (k,) posterior mean random-walk variances of the log volatilities.

  • n_draws – Total sampler iterations.

  • n_burn – Burn-in discarded.

  • thin – Post-burn thinning.

Private methods#

_from_fit(fit, model)

Assemble the public result from a raw fit and its specification.

_predictive_noise(draw, steps, rng)

Innovations under the draw's own forward-simulated volatility.

_replication_noise(draw, rng)

Innovations under the draw's own in-sample volatility path.

_replication_notes()

Remarks a predictive check carries about how this result replicates.

_summary_table()

Build the structured summary.
On this page