Kalman and forecasting!

[nimz]

Hey,

I know this has been posten before but i couldnt find a solution in that thread. First and foremost, I am using eview 6.
I have managed to get results for my state space model but now I want to forecast the value for my state. My model is very simple :

@signal bcb = sv1*ftseallshare + [var = exp(c(1))]

@state sv1 = sv1(-1) + [var = exp(c(2))]

i have used proc>forecast
forecast method >smoothed
Forecast output> states

but i dont know what forecast sample to select. I am using weekly data and this 12/27/2000 11/09/2008 appears in my forecast sample.

I have been trying to get the results using various sample options but i get a N/A in the for the period i have and no forecast...so basically my series is full with N/A with no forecasts

I want forecasts for the year ahead.

Kindly help as I have a submission soon and this is the last part of it and am stuck. :( :(

Thanks in adv
Nimz

[trubador]

First of all, you have to extend the size of your sample to include the forecast period (i.e. adjust your range not the sample). Second, if you want to forecast the "bcb", then you should have future values for "ftseallshare", since it is exogenous to the model. As a final remark, forecasted values of your state variable, sv1, will be equal to estimated final state figure, since it follows a random walk.

[nimz]

hey got it working.... :D :D :D :D :D :D
thanks a ton...God bless ya!!

[nimz]

well i became happie a lil sooner than i sud have..I understood how to do it. I have around 10 state space equations and i had already estimated them. However, for foreasting purpose i expanded by sample space so i start with the estimation of state space again...while i get the forcast results for a few equations i face two type of problems with some equations:
I get a singular covariance - coefficients are not unique warning. If I continue with the forecast i get 0.0000 in both the forecasted signal and the sv1 generated. I still get 0.0000 generated in some other equations in which the warning message is not coming.

In the earlier posts I read a solution to it maybe to make sure all values in the variable names "c" are zero. I have tried doinng that also but to no success

So basically i end up getting the forecasts for a few equations only.

Kindly help.

Thanks for the reply n help
Nimz