correctly applicating ARMA

[mozillaff]

Hello!

I've studied several textbooks about time series analysis and now I'm trying to implement some ARIMA models using eviews. The basic ideas of ARMA models seem quite simple and easy to understand, but nevertheless a lot of questions arise concerning the apropiate methodology for estimating an ARMA model.

I've got a series of quarterly gdp growth with 215 observation. I estimated an ARMA(5,3) model with
ls gdp c ar(1) ar(2) ar(3) ar(4) ar(5) ma(1) ma(2) ma(3).
ar(2), ar(3) and sar(4) have insignificant estimates while the rest of the estimates are significant. Now, what to do with the insignificant variables? Is it correct to exclude them and estimate a model with the significant variables only? That is:
ls gdp c ar(1) ar(5) ma(1) ma(2) ma(3)
But now ar(1) ar(5) are no longer significant. Eliminating the insignificant variables from the model feels wrong but is it correct to rely on the insignificant variables when I want to run a forecast?

Thank you very much for your help!

[mozillaff]

By the way, how do I perform an out of sample forecast??? I can't find any option for this in the forecast window! The Eviews help file remains silent on this topic.

[startz]

Actually, the help system discusses forecasting at length. But the simple answer is that the forecast window has a field named "Forecast sample." Fill it in with whatever sample period you would like to forecast.

[mozillaff]

Thanks for your advice. My problem was, that, first, one has to expand the sample with blank observations for the desired forecasting period which I hadn't done. Second Eviews can't run both in-sample and out-of-sample forecasts at the same time and I tried to do so. Now it works.