Page 1 of 1
Forecasting using Kalman Filter
Posted: Sat Jan 17, 2015 6:41 pm
by jmagomez
Dear Fellows,
I developed a model using the State Space Object with Kalman Filter, but I am not having success to forecast. What do I have to do?
I am sending the Eviews file and the object is called sskfvecmgasrealsubst. The variables are lgasolinac, lmassasalarialreal and lsubst. Please could you help me to forecast. What am I doing wrong?
Any doubt, please contact me.
Best Regards, Jmagomez.
Re: Forecasting using Kalman Filter
Posted: Thu Jan 22, 2015 3:31 pm
by jmagomez
Dear Fellows,
On what website could I have information about my doubt over forecasting using Kalman Filter? Please, I would like your help.
Best Regards, José.
Re: Forecasting using Kalman Filter
Posted: Fri Jan 23, 2015 12:02 am
by EViews Glenn
Have you read the discussion of forecasting using the state space object in the manual? If so, do you have specific questions regarding the material?
Re: Forecasting using Kalman Filter
Posted: Fri Jan 23, 2015 10:06 am
by jmagomez
Hello Glenn, thanks for your reply. I read the Eviews manual , the Kalman Filter topic, four times. My doubt is:
I developed a simple model named sskfvecmgasolinarealsubst and I am trying to forecast for one year ahead from 2013m07 until 201406 and make an out of sample analysis. I tried all different specifications: Dynamic, Smoothed and N-period ahead, but Eviews didn't return any answer and it didn't write where the problem is. Could you help me? You change the file.
Best Regards, José.
Re: Forecasting using Kalman Filter
Posted: Fri Jan 23, 2015 12:08 pm
by jmagomez
Dear Glenn,
I would like to add that I am using a VECM. Maybe Eviews can not forecast VECM using Kalman Filter.
Best Regards, José.
Re: Forecasting using Kalman Filter
Posted: Fri Jan 23, 2015 5:26 pm
by EViews Glenn
As we note in the manual, if you are forecasting in a state space with a dependent variable as an expression, you have to provide names since the wildcard can't create a name using the existing dependent. Just enter one or more name.
Re: Forecasting using Kalman Filter
Posted: Sat Jan 24, 2015 7:40 am
by jmagomez
Thanks a lot, Glenn. I had success on forecasting using your tip. I didn't do this before, because there is an expression at page 744 - log(passenger) - that says the referenced equation would be correct. I give a suggestion to allow forecasting expressions at state space models at the newest versions from Eviews.
Best Regards, José.
Re: Forecasting using Kalman Filter
Posted: Mon Jan 26, 2015 2:21 pm
by EViews Glenn
Estimating the equation with an expression is not a problem. Just when you forecast the signals, you have to provide explicit names.