Hi Abdelrazzaq,
-For your first question, you can just uncomment line 50 ( and comment
in line 49) of the Forecast.ox example .ox file
(OxMetrics7\ox\packages\Garch\samples\Forecast.ox),
in doing so you will compare the forecast conditional variance to the
(observed) squared demeaned returns with is a common proxy for thre
"true" volatility if you don't have
the realized volatility (ie: if you don't have intraday data).
Ex : so yo should get :
decl Hfor =
garchobj.GetVar("REALVOLA")[T+step-1:T+step-1+rows(yfor)-1]; //
Realized volatility
// decl Hfor = (yfor - meanc(yfor)).^2; // Squared
returns (in deviation)
instead of :
// decl Hfor =
garchobj.GetVar("REALVOLA")[T+step-1:T+step-1+rows(yfor)-1]; //
Realized volatility
decl Hfor = (yfor - meanc(yfor)).^2; // Squared returns
(in deviation)
As important output i think you are interested with the line 59 :
garchobj.MZ(Hfor, forc[][1], number_of_forecasts); which perform the
Mincer-Zarnowitz regression. (see g@rch Help)
-For the second question, i don't think there is CGARCH models in G@rch
but if you are interested in the long memory feature of CGARCH you can
use long memory models provided
in the package (FIGARCH,FIAPARCH..).
I hope this can help you,
Bests,
Malick
Le 29/07/2014 00:31, Abdelrazzaq Alrababa'A a écrit :
> Hi there;
> I am trying to forecast the stock retrun volatility using the G@RCH 7 package and just have two questions:
>
> 1- how to change the true volatility proxy for assessing the forecasting performance of GARCH models. , which proxy the G@RCH use? I need to use the perfect unbiased proxy for Pagan & Schewart (1990) which it is the squared of residuals from the conditional mean equation? How to write the code for that?
> 2- How to estimate and forecast the CGARCH model?
>
> It is urgent please advise and if there is a ready written code, then I will appreciate.
|