Page 1 of 2

Problem in Automatical ARIMA Forecasting

Posted: Sat Jun 20, 2015 11:12 am
by economist79
My sample years are 1972-2013 and my range years are 1972-2025 so I want to make an automatic ARIMA forecast for my dependend variable. I made the forecast and I can see the actual values for my sample years (1972-2013) but I can not see the forecast values for my ex-ante years (2014-2025). Can you help me with my problem, thanks.

Re: Problem in Automatical ARIMA Forecasting

Posted: Sat Jun 20, 2015 1:34 pm
by EViews Gareth
How many periods did you forecast over?

Re: Problem in Automatical ARIMA Forecasting

Posted: Sat Jun 20, 2015 2:08 pm
by economist79
I forecasted 12 periods for the years 2014-2025. Eviews9 selected the best ARIMA model for forecast, and so everything seems normal, only I can not see the values of forecasted years neither as graph nor as number.

Re: Problem in Automatical ARIMA Forecasting

Posted: Sat Jun 20, 2015 8:09 pm
by EViews Gareth
Can you provide the workfile, and step by step instructions on what you did?

Re: Problem in Automatical ARIMA Forecasting

Posted: Sat Jun 20, 2015 10:53 pm
by economist79
eviewsfile.pdf
Eviewsfile
(42.43 KiB) Downloaded 1014 times
At the beginning I composed a workfile and added my variables for the years 1972-2013, then I set the sample as 1972-2013 and then I increased my range up to 2025 by adding the command (expand 1972 2025). After these steps I double clicked my dependend varible (actuary) and clicked proc menu and automatic ARIMA forecasting. At the automatic ARIMA forecasting window there are two parts. One is specification part and other is options part as you know. At the specification part I only added constant value and my other independend variables as regressors, and after that ,at the options part I clicked the Forecast comparison graph, ARMA criteria table and ARMA criteria graph. These are, step by step, what I have done so far.

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 10:15 am
by EViews Gareth
Do you have data on the independent variables over the forecast period?

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 10:46 am
by economist79
I have no data on independent variables over the forecast period cause i am trying to make ex-ante estimations for the years 2014-2025, is this not possible via eviews9?

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 10:48 am
by startz
Nothing to do with EViews. If you have independent regressors then you need values for them during the forecast period. You either need to know them or to forecast them.

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 10:49 am
by economist79
You're surely right, i tried to forecast my independend variable without adding the regressors and now i can see the forecasted values. Can i make the same forecasts for my independend variables and then add these values ass regressors for forecasting my dependend varibable?

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 10:51 am
by startz
You might want to look at the VAR feature.

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 10:55 am
by economist79
Does Var feature has automatic forecast property too?

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 11:01 am
by EViews Gareth
No, you'll have to choose the number of lags manually.

Re: Problem in Automatical ARIMA Forecasting

Posted: Sun Jun 21, 2015 11:08 am
by economist79
Thank you guys for giving me information about ARIMA forecast, i will try both Arima and Var model and select the best model proper to the literature.

Re: Problem in Automatical AR and MA Forecasting

Posted: Wed Jun 24, 2015 2:22 pm
by Hasna
Hi,

I have two models AR(1) and MA(1), I was able to make forecast for one period ahead but I want to make forecast for at least 3 months ahead.
Since there is no actual yet for Y+1 period therefore for the year Y+2 I cannot make any ax-ante forecast because to estimate the equation I need error et-1 for MA(1) where et is the error of Yt period forecast. Likewise AR(1) i dont have the actual for Y+1 period and to estimate forecast for the time Y+2 i need actual for the previous period. In that case should I use Y+1 forecast value to make forecast for Y+2 same as error for the MA(1) model?

Thanks/Hasna

Re: Problem in Automatical ARIMA Forecasting

Posted: Wed Jun 24, 2015 2:28 pm
by EViews Gareth
Just do a dynamic forecast.