HierarchicalBVARResult.sigma_u

HierarchicalBVARResult.sigma_u: NDArray[float64]

(k, k) exact posterior mean of the innovation covariance.

\(\bar S / (\bar\nu - k - 1)\), the mean of the inverse-Wishart posterior, not the average of sigma_draws, which converges to it as the draw count grows. It is the covariance simulate() uses at the posterior mean (draw=None).