Page 1 of 1
Forecasting a estimated equation
Posted: Tue Nov 11, 2008 5:40 pm
by LightMooCow
So, I know there is a lot of help on this but I am still having problems with this.
I have data for two variables: ff and ip from 1989m1 to 2008m1. The goal in mind is to forecast from 2008m2 to 2010m2.
So, I estimate the model dip c dip(-1) dip(-2) ff(-1) ff(-2) where dip = log(ip)-log(ip(-1)).
When I try to forecast from 2008m2 to 2010m2, it says that I need to go to proc structure/resize the page. So I then change the range from 1989m1 to 2008m1 --------> 1989m1 to 2010m2. However, when I try to forecast 2008m2 2010m2, it says that there is not enough observations.
What I don't understand is that I have the data and I have already estimated the equation via OLS. Can someone shed some light on this matter please? :)
Re: Forecasting a estimated equation
Posted: Tue Nov 11, 2008 5:47 pm
by EViews Gareth
Did you do a static forecast or a dynamic forecast?
Re: Forecasting a estimated equation
Posted: Tue Nov 11, 2008 5:48 pm
by LightMooCow
I tried both and I got the same error :'(
Re: Forecasting a estimated equation
Posted: Tue Nov 11, 2008 5:52 pm
by EViews Gareth
A static won't work, but a dynamic should.
From your equation click on Proc->Make Regressor Group. Then check that you have data for all variables (other than the dependent) for the first period of the forecast sample. If not, then that's your problem.
Re: Forecasting a estimated equation
Posted: Tue Nov 11, 2008 6:02 pm
by LightMooCow
The data goes up to 2008m09
So, what I see are:
obs dip dip(-1) dip(-2) ff(-1) ff(-2)
2008m09 ### #### #### #### #####
2008m10 na na na na na
. . . . . .
. . . . . .
. . . . . .
So, when I do a dynamic forecast from 2008m09 2010m10, the forecast sample is 2008m09 2010m10 and the adjusted sample is 2008m09 2008m10. Included observation is 1. Any ideas?
By the way, I really appreciate your very, very quick responses. Thanks
Edit: It didn't keep the spaces, I tried to show in the awesome table I made... :'(
Re: Forecasting a estimated equation
Posted: Tue Nov 11, 2008 6:11 pm
by EViews Gareth
What you are describing is odd.
If dip exists for 2008m09 (I presume the ### indicate that it does exist), then, dip(-1) should exist for 2008m10.
Similarly if ff(-1) exists for 2008m09, then ff(-2) should exist for 2008m10.
Re: Forecasting a estimated equation
Posted: Tue Nov 11, 2008 6:28 pm
by LightMooCow
Yeah, it is odd. It seems that I was not the only one having problems on this though. Other people doing the Problem set are having problems too
I guess since I have the estimated model, I can just use excel to produce the forecast for the next 24 months.
Also, I was able to obtain the forecast using proc --> make model (it actually is a VAR(2)). However, I thought I would be able to obtain the SEs or SDs, but in the other thread, nope
I'll probably look into stata... maybe... if I have the time...
Anyway, I really appreciate your help. Thank you so much!!