What will change in VAR model if I will introduce robust estimators of covariance variance (Newey-West)? Will only the interpretations change and the properties of the model remain the same? Or maybe nothing will change in the properties or interpretation of the model, but it will be resistant to autocorrelation? 📷
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
The VAR is supposed to consider the auto correlation, if you suspect the correlation is still there test using more lags. Robust estimators (HAC and variants) usually do not change the estimated coefficients just the estimation of the error or confidence intervals. The interpretation is the same. There is a related answer and a workaround here https://stats.stackexchange.com/a/390141.
