I don't know about arima, but I'm doing the same thing with other regressions and the gyst of how you do it is if you have the formula Y=XB and you want to forecast Y forward you need observations (or estimates) of the X's for the time period you want, then some way to set the sample space for your forecast. What I did was I appended the observations of the X's for which there was no Y to the endo fthe data set, then created a dummy variable that was 1 when there was a Y and 0 otherwise (I did this part in excel). then I just used smpl dummyvar --dummy (I can't remember if this is the exact switch, look in the manual) then fcast in the standard form in the manual. (ie fcast fit2). Momentarily scanning the ref pdf to find the fcast fit2 command above I noticed that you don't even have to use the dummy variable thing I mentioned above, you can set the sample to what you want right in the fcast command.

Hope this helps


On 6/20/07, walgula@wp.pl <walgula@wp.pl > wrote:
Hello, I'm trying to use an a raima model to forecasting.
How to write a script that forecasts forwards some variable?
I've dataset/file that includes "X" observations in time and i want to forecast this observation in time hour+1 or hour+2 forewards.
How to do it? I know that I should to use "fcast" or "fcasterr" functions, but i don't know how to get time parameters.
Please, help me.
Best regards, walgula

Gretl-users mailing list