Page 1 of 1

predictions

Posted: Sun Apr 03, 2016 3:58 am
by jmmbelbute
Dear all
I need to build a reference scenario for a variable using an ARFIMA univariate model, estimated for the period 1900-2014.
I’ve already got the in-Sample and out-of-sample forecasts for this period but the reference scenario covers predictions from 2015 to 2050.
Until now I’ve been unable to understand how can I do it using Eviews... :(
Therefore, I would be grateful if I could have any help from you to solve this "problem" :)
Thank you
Regards

Re: predictions

Posted: Sun Apr 03, 2016 7:07 am
by EViews Gareth
It isn't clear what you're asking.

Re: predictions

Posted: Sun Apr 03, 2016 8:31 am
by jmmbelbute
Thank you for yor reply
For all that matters, I need to get forecasts of x(t) between, say, 2000 and 2100, from a pre-estimated ARFIMA model on x(t), using eviews
It should be noted that I have already estimated the ARFIMA model in e-views.
regards

Re: predictions

Posted: Sun Apr 03, 2016 8:36 am
by jmmbelbute
...
Please, note also that my sample period for x (t) goes from 1900 to 2014
JMMB

Re: predictions

Posted: Sun Apr 03, 2016 3:46 pm
by EViews Gareth
Isn't that just a forecast? Just hit the forecast button and change the sample to be whatever you want it to be

Re: predictions

Posted: Mon Apr 04, 2016 1:49 am
by jmmbelbute
That's what I thought...!
Anyway, I ‘ve already done it (I've even expanded the workfile range until 2100) but it still doesn’t work. I'm still getting the in-sample predictions/forecasts (1900-2014), and I don’t know what's missing? Can I send you my workfile?

Re: predictions

Posted: Mon Apr 04, 2016 4:28 am
by EViews Gareth
You can attach it to a post here

Re: predictions

Posted: Wed Apr 06, 2016 1:49 am
by jmmbelbute
Here is the file with one ARFIMA models estimated in E-Views 9 (64 Bit).
I plan to get forecasts/predictions between 2015 and 2100. I’ve already extended the workfile range until 2100.
I'd be extremely grateful for your help (I’m able to get predictions (out-of sample) using stata but the stata’s arfima module does not allow for the use of dummies accounting for the presence of breaks, which are crucial in my sample)
JMMB

Re: predictions

Posted: Wed Apr 06, 2016 4:40 am
by trubador
You should have/generate out-of-sample values for each independent variable in your model (i.e. T and T46).