Automatic ARMA Forecasting

[Bablowski]

Hello,

I used Eviews 9.5 for constructing ARMA model with new function Automatic ARMA Forecasting. But I don´t undestrand the following:

Eviews gave me the automatically computed model ARMA(4,3). The thing I don´t understand is that the coefficient MA(3) is not statistically significant (according to p-value) but I think it should be statistically significant if the model is the best. Could you please help me to understand how to evaluate the results?

I submit the model Eviews gave me.

Thanks a lot for any help.

[trubador]

Information criteria are used in model selection, not the individual p-values. Also your output reads: "Convergence not achieved after 500 iterations". It is difficult to pinpoint the source of the problem without seeing the actual workfile/data.