GibbsBVAR.prior_replications#

GibbsBVAR.prior_replications(n_replications=200, *, seed=None)[source]#

Data sets the independent Normal-Wishart prior generates.

Each replication draws Sigma from its inverse-Wishart prior and every coefficient from its own Gaussian, independently of Sigma, then simulates the sample from the observed presample.

Parameters:

  • n_replications (int) – Replicated data sets.

  • seed (int | Generator | None) – Seed or generator.

Return type:

NDArray[double]

Returns:

An array of shape (n_replications, nobs, k) aligned with observed.

Raises:

SpecificationError – If the count is not positive, or the prior is adaptive or states dummy observations. An adaptive prior’s coefficient variances are latent states with no marginal prior to draw from in closed form; dummy rows under an independent prior couple the coefficients to the covariance through pseudo-data, so the effective prior has no direct sampler. The conjugate BVAR folds dummy rows into its prior exactly and replicates them.