[R-SIG-Finance] Fwd: Problems with rugarch package

[Camila Villegas]

Greetings


My name is Camila Villegas, I  write to you about a fault that I found in
the rugarch package and that I would like to discuss with you.

For my master's research, I conducted simulations with the rugarch package
to evaluate its operation with a TGARCH model (1,1).

First I analyzed the convergence of the estimated parameters and they
presented no problems. plus the mean square error decreased as the sample
size increased.

The validation of the modeling results (obtained by the "ugarchfit"
function) I carried out by analyzing the residues generated by the
"residuals" function. The simulated residues originally had zero mean and
variance one, however, in the 10,000 simulations that I developed, the
values of the variances of the residues generated by the "ugarchfit"
function were significantly different from one.

On the contrary, when calculating the residues "manually" and only using
the parameters estimated by the package, we obtained means close to zero
and variances close to one.

With this last point, I observed a flaw in the residues delivered by the
rugarch package for a TGARCH process (1,1), since they do not have the same
variance with which they were simulated.

Thanking your observations,

Kind Regards.

Camila Villegas Orellana

	[[alternative HTML version deleted]]