statsmodels.tsa.vector_ar.var_model.VARResults.forecast_cov¶

VARResults.forecast_cov(steps=1, method='mse')[source]¶

Compute forecast covariance matrices for desired number of steps

Parameters:¶

stepsint

Returns:¶

covsndarray (steps x k x k)

Notes

$${\mathrm{\Sigma }}_{\hat{y}}(h)={\mathrm{\Sigma }}_y(h)+\mathrm{\Omega }(h)/T$$

Ref: Lütkepohl pp. 96-97