Page 1 of 1

kalman filter and forecasting "missing values"

Posted: Mon Jul 19, 2010 2:06 pm
by maragloria
Hi there,

I have estimated a state-space model for the period 1981Q1 to 2007Q4.

Now, I would like to use the estimates to forecast until 2010Q2 (the signal variables are available only until 2007Q4). I read in some textbooks that the Kalman filter can estimate “missing values” (which in my case would be from 2008Q1 to 2010Q2). Does Eviews 7 perform such forecasts?

Thanks very much for your help.

Re: kalman filter and forecasting "missing values"

Posted: Mon Jul 19, 2010 2:57 pm
by EViews Glenn
Use the forecast proc.

Re: kalman filter and forecasting "missing values"

Posted: Tue Jul 20, 2010 6:11 am
by maragloria
I did use Proc Forecast but I am getting "NA" for the period 2008q1-2010Q2. Could you please tell me what are the steps to follow? I've tried a few things already but I am getting no results.
Thanks for your reply.

Re: kalman filter and forecasting "missing values"

Posted: Tue Jul 20, 2010 8:42 am
by maragloria
Hi Glen,

Could you please help me with the forecast?

I am still getting NA with proc forecast.

Do I have to change the sample and range period in the workfile? Itried that and I re-estimated the model but I still got nothing.

Thanks,

Mara

Re: kalman filter and forecasting "missing values"

Posted: Tue Jul 20, 2010 8:59 am
by EViews Glenn
That's a hard question to answer without seeing your model. A likely possibility is that you have non-state variables in your signal equation which do not have values for the out-of-sample period.

Re: kalman filter and forecasting "missing values"

Posted: Tue Jul 20, 2010 9:17 am
by maragloria
Yes, I do have 4 pre-determined variables in the state specification.
I assume I need to supply values for those variables for the period I want to forecast (UPS!). I'll do that now and see what happens.

Thanks very much for your reply.