EViews.com

Dear all,
I am trying to estimate an arima model. I was close to end but a lecturer warned me about the sum of the absolute values of AR and MA coefficients have to be smaller than 1. i haven't read anything on several books like this and i wonder if there is such a restriction. Can anybody clarify me please?

Thanks

You probably have overlooked this while studying. Every econometrics/time series textbook mentions about this restriction, since it is a very crucial fact in terms of stationarity (or invertibility). Please refer to your sources and carefully go over the subject again.

thanks trubador. i got it . then in that case i have to drop some of the ar or ma processes even though they are significant; untill the absolute total of the coefficients becomes smaller than 1. Am i right ?

Not necessarily. You should follow the Box-Jenkins methodology step by step and try to find out what went wrong in your model identification stage. Since the methodology assumes that the time series is stationary, you should also carefully assess whether your series violates this assumption before going further.

Hello
If you know french, I send you a file about Box-Jenkins