GibbsBVARResult._lag_diagnostic#

GibbsBVARResult._lag_diagnostic()[source]#

The lag-block rows of shrinkage as a (p, k, k) stack.

Skips the deterministic rows and transposes each lag block so that entry [lag, i, j] refers to equation i and regressor j, matching coefficients.

Return type:

NDArray[double]

Returns:

A (p, k, k) stack.

Example

>>> import numpy as np
>>> from cultivars.bayes.priors import HorseshoePrior
>>> y = np.random.default_rng(0).standard_normal((150, 2))
>>> res = GibbsBVAR(y, order=2, prior=HorseshoePrior()).fit(
...     n_draws=60, n_burn=20, seed=0
... )
>>> res.shrinkage.shape, res._lag_diagnostic().shape
((5, 2), (2, 2, 2))
>>> bool(np.allclose(res._lag_diagnostic()[0], res.shrinkage[1:3].T))
True