VECMResult.fevd#

VECMResult.fevd(horizon=20)[source]#

Forecast-error variance decomposition.

Parameters:

horizon (int) – Largest lead to return.

Return type:

NDArray[double]

Returns:

An array of shape (horizon + 1, k, k) whose entry [h, i, j] is the share of variable i’s h + 1-step forecast-error variance attributable to shock j. Rows sum to one by construction, so a row that does not is a bug rather than a finding.

Note

Inherits the ordering dependence of irf() in full, and more visibly: at h = 0 the decomposition is exactly triangular, so the first variable is always attributed 100% of its own impact variance purely because of where it sits in names.