I am looking for the error of a GARCH-Model.
More precisely:
I estimated an AR(1) with GARCH(1,1) errors for stock market data. Now I wanted to plot the residuals of the model.
To my surprise they still showed the same conditional heteroskedasticity as without the GARCH-modelling.
So I guess these are the Residuals of the Mean Equation, i.e. the AR(1).
But I am looking for the error of the GARCH-Equation.
When EViews conducts ARCH-Tests, it will use "WGTRESID" which I assume is what I am looking for. How can I plot these?
Actually I only want these residuals for teaching / explaining to students. But it should be possible to somehow get them, shouldn't it? It seems to me like these residuals should also be shown after estimating a GARCH Model when clicking on Graph Residuals. Or am I interpreting the GARCH approach a bit wrongly? This is, does it actually make sense to look at them? I am not sure
I guess it does because they are the ones tested in the ARCH-Test so they have some meaning. The human eye might catch some left over structure that the ARCH-Test doesn't find anymore because precisely all the ARCH-Effects have been removed. The residuals might still not be white noise, however.Thanks alot!