Forecasting

[j52]

Hey,

I would like to forecast an equation with independed lagged variables (and arma terms if possible). However, when I'm using the statistic forecast, I only get a forecast for one period ahead even though my lagged variables are (-13). (Using differenced series I get a longer forecast, but using the d(y) term I just get one month.)

And regarding the arma terms, is it possible to integrate them into the forecast with out loosing the forecast hoizon (13 month ahead)? Maybe with a stepwise forecast!?

Would be great if somebody can help me with that!

Thanks a lot.

[EViews Gareth]

Do you have data for the forecast period?

[j52]

no, but I'm not sure what you mean. I have monthly data, untill 06/2013 and what to create a forecast for the next 6 month or next year. I know that with the statistic forecast I usually get one period forecast, but I figured since my independed variables are lagged it shoud be possible to create a longer forecast, since the independent variable from 06/2013 has for example an effect on the dependent variable in 12/2013!? Or do you mean, that it's not possible if I do an out-of sample forecast? But the wired thing is that with the defined series for the first difference it works!?...

[EViews Gareth]

To perform a static forecast, you need data for the regressors.

[j52]

yes, I know. But shouldn't it work when I have leading independent variables?

[EViews Gareth]

No.

[John88]

Hi. I would like to ask you how to perform multi-step ahead forecasting on eviews. I use the GARCH family models for the estimations, but do not know how to perform the multiple days ahead forecasting, like 1, 2 ,7 ,20 days ahead and so on. I read that it should be done by dynamic forecasting button, but when i press this button it just creates the series for the forecasting horizon, where all the numbers are the same. For example, if my out-of sample forecasting consists of 1000 observations, the new series emerging after pressing this button will have 1000 identical numbers. So what is wrong here? I think the numbers should be different, am I right?

[EViews Gareth]

Do you have regressors in the mean equation?

[John88]

EViews Gareth, thanks for your reply, yep I have AR(1) in the mean equation, but even without it the result is the same

[EViews Gareth]

Without an exogenous variable, the forecast will be constant.

[John88]

but what i want to know is that if i choose dynamic forecast and use the daily data, and the horizon for the forecast is 60 days (i mean the forecast includes 60 daily observations), does it mean that 60 days-ahead forecast will be created ? so, if i want to do n days ahead forecast , i need to choose n days observation for the out-of sample forecast and implement dynamic forecasting on eviews, am i right?

[EViews Gareth]

Yes.

[ah09185]

Hello, I have an in sample from 1st april 2002 till 30 march 2011 and an out of sample from 1 april 2011 til 30 march 2012. when I forecast using dynamic method the variance and covariance proportions turn out to be NA. Also when I use static methos and generate the new variable lets say FORECAST all the out-of sample values are the same. What IS IT WRONG?

thank you in advance

[EViews Gareth]

Do you have regressors in the variance equation?

[ah09185]

no! i am testing for garch(1,1) tgarch(1,1,1) and egarch(1,1). i estimated the models by including the loged daily returns.
thank you in advance