The RATS Software Forum

Ah, i saw the problem.
Thanks

Dear Tom, Please help me to see and to fix the problem at this out of sample forecast: CALENDAR(M) 2009:06 allocate 2012:07 open data forecast.rat data(format=rat,org=columns) Boxjenk(regressors,ma=2,define=equation) ENDO 2009:01 2012:06 # exo @uforeerrors inf forecast 2009:6 2012:06 UFORECAST(equat...

Thank you very much, I've run the model ARIMA(0,1,2) with the armadlm and DLM estimate. Il works better than UC model with Veta_T=0.0. This result will be take as an exogenous variable of the complete model which i have asked you in the second post. I hope for a good result. Many thanks for your adv...

Oh,
Thanks a lot for your advices.
I review the series and try to find the better models.
Can you suggest me one? Please.

oh, I just re-estimate the model without Z option (no fixed trend mu). and I attached here the full program and the data set It still has the problem in which the last values (at 2013:03) are too larges for both trend and cycle component. So,Please tell me what solution could i take to solve this pr...

Dear Tom, Please help me to see why the last values of the estimated states are very very large in comparison with the others values of the states in this case. Many thanks for your helps. * * Create the matrix dec frml[rect] a dec frml[vect] z dec frml[symm] sw * nonlin mu Veta_T ph1 ph2 Veta_C Vet...

Thank you very much. It works.

Dear Tom, I'm trying to estimate a univariate state space model including one exogenous variable in the state and one exogenous variable in the observation equation. I got problem at the DLM estimation with the error : ## MAT2. Matrices with Dimensions 1 x 5 and 3 x 1 Involved in ~~ Operation So, pl...

Many thanks for your help.

Dear Tom, I'm trying to replicate the KalMan filter of Stock and Watson (1991) and I would like to extract the Kalman Gain. I have added the option: Gain = Gain into DLM * dlm(start=%%DLMSetup(),y=yf,sw=sw,c=c,a=a,f=f,$ presample=ergodic,type=filter,$ pmeth=simplex,piters=10,Gain=Gain,method=bfgs) 1...