HierarchicalBVAR._prior_at#

HierarchicalBVAR._prior_at(theta)[source]#

The prior a hyperparameter vector names.

The vector is always three long – (lambda, mu, delta) in the paper’s parameterization – with inactive dummies simply not composed in. The dummy priors take a tightness, so mu and delta enter as their reciprocals.

Parameters:

theta (NDArray[double]) – (lambda, mu, delta), all positive.

Return type:

_Prior

Returns:

The composed prior.

Example

>>> import numpy as np
>>> panel = np.random.default_rng(0).standard_normal((100, 2))
>>> model = HierarchicalBVAR(panel, order=1)
>>> model._prior_at(np.array([0.5, 2.0, 4.0]))._label()
'niw(l1=0.5, l3=1, l4=100) + soc(0.5) + dio(0.25)'