[R-SIG-Finance] Hessian in GARCH-Models (rugarch-package)

[Lin23]

There are two options for the robust covariance in the  'robustvcv' -code.
The first way to calculate the robust covariance is the sandwich-estimator :
B = cov(scores)
vcv = (Ainv%*%B%*%Ainv)/n
The second way uses the NeweyWest-Weights 
B = neweywestcv(scores, nlag)
vcv = (Ainv%*%B%*%Ainv)/n
When do I use the sandwich-estimator and when do I use the
NeweyWest-Weights?
In the literature on GARCH-models you often find "p-values based on
Bollerslev and Wooldridge
(1992) robust standard errors" or " p-values based on Bollerslev and
Wooldridge
(1992) robust standard errors with six lags". I guess in this case they use
the first option.
My model takes into account first-order and third-order autocorrelation in
stock returns, a structural change in the autoregressive structure,
international interdependence of the U.S. stock market, and a
day-of-the-week-effect.
I guess I should use the first option to calculate my robust covariance.
I adopt your "box.test".  Do I have to change des df (degrees of freedom) in
the Ljung-Box-test? By default they are zero. I use GJR-GARCH(1,1) and a
complex mean-equation (see above). If I am not mistaken there is no need to
change the default-setting for the box.test.
Thank you

--
View this message in context: http://r.789695.n4.nabble.com/Hessian-in-GARCH-Models-rugarch-package-tp4371133p4379242.html
Sent from the Rmetrics mailing list archive at Nabble.com.