Page 1 of 1

forecast of nan(series,0) with AR model

Posted: Tue Sep 14, 2010 1:19 pm
by minijim
Hi,

I'd like to make a forecast for a given series y, which contains some NA values.
I don't want to substitute these NA values with 0 in the series, but I thought that i could use @nan(y,0) instead of y. Then I perform a AR(1) model, as follows :

Code: Select all

equation eq.ls @nan(y,0) c ar(1)
The estimation output seems to be ok. Nevertheless, when I try to make a forecast of this equation,

Code: Select all

eq.forecast yf
the following error message appears : "Unable to compute due to missing data".

Could anyone tell me where my mistake is or what is the cause for this message ?
Thanks in advance