statsmodels.tsa.vector_ar.svar_model.SVARResults.forecast_cov#

SVARResults.forecast_cov(steps=1, method='mse')#

Compute forecast covariance matrices for desired number of steps

Parameters:
stepsint

Number of steps ahead to compute forecast covariances for.

method{“mse”, “auto”}, default “mse”

If “mse”, use the forecast MSE, ignoring parameter uncertainty. If “auto”, also take parameter uncertainty into account by adding the forecast error covariance due to parameter uncertainty; this is currently only supported if there is no exogenous data and the trend is one of “n” or “c”.

Returns:
covsndarray (steps x k x k)

Notes

\[\Sigma_{\hat y}(h) = \Sigma_y(h) + \Omega(h) / T\]

Ref: Lütkepohl pp. 96-97