Hi every one
?
I would like to compute MSE on out of sample forecast from GARCH model. I used rgrach package to obtain the estimated variance (sigma2.hat). then I run an out of sample forecast as
forc = ugarchforecast(fit.garch, n.ahead=100)
this will give me 100 observation of sigma.forc( the forecasted values)
to obtain the MSE theoretically is
MSE = 1/h+1 { sum(sigma2.hat ? sigma2.forc)^2}
What confuse me is that the length of sigma2.hat actually is equal to my data length (2000 observation), while the length of sigma2.forc is only 100.
?
Can someone kindly enlighten me about this.
?
Sincerely,
?Shamiri
MSE from GARCH forecast
3 messages · ahmed_shamiri at yahoo.com, Alexios Ghalanos
I am not sure I understand your question: - the estimated variance is supposed to equal the length of the data, so length(sigma2.hat)==length(data) - the length of the sigma2.forc was made with ugarchforecast(fit.garch, n.ahead=100), i.e. you requested 100 forecast values. HTH -Alexios
ahmed_shamiri at yahoo.com wrote:
Hi every one
I would like to compute MSE on out of sample forecast from GARCH model. I used rgrach package to obtain the estimated variance (sigma2.hat). then I run an out of sample forecast as
forc = ugarchforecast(fit.garch, n.ahead=100)
this will give me 100 observation of sigma.forc( the forecasted values)
to obtain the MSE theoretically is
MSE = 1/h+1 { sum(sigma2.hat ? sigma2.forc)^2}
What confuse me is that the length of sigma2.hat actually is equal to my data length (2000 observation), while the length of sigma2.forc is only 100.
Can someone kindly enlighten me about this.
Sincerely,
Shamiri
_______________________________________________ R-SIG-Finance at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-finance -- Subscriber-posting only. -- If you want to post, subscribe first.
I have now preliminarily added some forecast performance measures (incl. mse, rmse etc) to the package which should be available to download from r-forge during the next build cycle (see the example in the ugarchforecast help). regards, Alexios
ahmed_shamiri at yahoo.com wrote:
Hi every one
I would like to compute MSE on out of sample forecast from GARCH model. I used rgrach package to obtain the estimated variance (sigma2.hat). then I run an out of sample forecast as
forc = ugarchforecast(fit.garch, n.ahead=100)
this will give me 100 observation of sigma.forc( the forecasted values)
to obtain the MSE theoretically is
MSE = 1/h+1 { sum(sigma2.hat ? sigma2.forc)^2}
What confuse me is that the length of sigma2.hat actually is equal to my data length (2000 observation), while the length of sigma2.forc is only 100.
Can someone kindly enlighten me about this.
Sincerely,
Shamiri
_______________________________________________ R-SIG-Finance at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-finance -- Subscriber-posting only. -- If you want to post, subscribe first.