Em 4 de agotos de 2011 Ignacio Diaz-Emparanza escreveu:

If you want to forecast out of the sample you should add the corresponding "space" for the new observations before using 'fcast', the command is

dataset addobs 10

(for adding 10 new obs)

Dear Ignacio,

Thanks a lot for your answer, but I think this is not the problem, once I'm estimating my models using a subsample. Please take a look at my script:

<hansl>
open australia.gdt
smpl ; 1989:4

var 4 E PAU PUS --quiet
fcast --out-of-sample
series E_hat = $fcast[,1]

smpl --full
print E_hat --byobs
</hansl>

The "australia" data ends in january/1991 and the sample used for the estimation ends in april/1984.

Best,
Henrique C. de Andrade
Doutorando em Economia Aplicada
Universidade Federal do Rio Grande do Sul
www.ufrgs.br/ppge