StudentBVARResult.irf#

StudentBVARResult.irf(horizon=20, *, orthogonalized=True, cumulative=False, shocks=None, responses=None)[source]#

Posterior mean impulse responses.

The mean of each draw’s response, not the response of the mean draw – the distinction matters because propagation is nonlinear in the coefficients. Orthogonalization applies each draw’s own Cholesky factor, so the ordering of names is an identifying assumption, as for any recursive scheme.

Parameters:

  • horizon (int) – Largest lead to return.

  • orthogonalized (bool) – Rotate each draw by its own Cholesky factor.

  • cumulative (bool) – Return running sums.

  • shocks (Sequence[str] | None) – Shock variables to keep, in the order given; all when omitted. Subsetting changes what is stored, not what is computed, so it bounds memory for large systems.

  • responses (Sequence[str] | None) – Responding variables to keep; all when omitted.

Return type:

NDArray[double]

Returns:

An array of shape (horizon + 1, n_responses, n_shocks); entry [h, i, j] is the response of variable i at lead h to shock j.

Raises:

SpecificationError – If horizon is negative or a name is unknown.