EViews.com

got the answer for my topic, Show next message below thanks

Looks like you are just estimating the mean equation as arma(1,1) and the variance equation as Garch(1,1) by maximum likelihood with the mean equation assumed to follow a t-distribution.

Garch(1,1) is pretty standard and the distributional assumption doesn't sound outrageous.

Your forecast specification can either by static or dynamic. I'd look at that part of the user manual.

Thanks for help

I have another question,

i get that graph :


What does it means ?

thanks

Looks like a graph of the forecasted (static?) series with +-2 standard error forecast confidence intervals.

The static forecasts are a series of 1-step ahead forecasts.

thanks for answer,

I don't know how to proceed. i have my garch equation : h(t) = c + 0,15 ut-1^2 + 0,88h(t-1)
so io should estimate the sigma(t+1).

But i can't estimate ht(t+2) (t+3) ... cause i don't have ut+1 ut+2 ...
Forecasting :
I have a sample range of 504 observations, so i resize the sample 1 - 300 i use garch modele to make an estimation of the conditional variance equation.

so i got this output :




Then i try to forecast the conditional variance. on the sample 300-504 like i have estimate on 1-300 i could call it "ex-post" forecasting



and i get this conditional variance.



So now i estimated with garch modele on the sample 300-504 and i get this conditional variance :

Is that the good procedure ?

I means my modele look like to be good no ?

if i have understood I cant forecast out of the sample ? or only for 505 ? cause after i don't have enought resid(t)^2 ?



So i can forecast the conditional variance t+1 ? and if i do it every day i could forecast forever

thanks you