Lag order ARCH(q)

[sonja509]

Hi there -

I'm trying to estimate an ARCH(q)-model for a time series of approximately 2000 observations and decided to use the Schwarz-Bayes information criterion to determine the lag order q.

Now my question:
The lowest SIC I obtained was for the ARCH(9)-model. However I have one insignificant parameter in this model, namely a1.
ARCH(1)-ARCH(8) have all only significant parameters and the information criteria decreases gradually with the lag order.

Now for my further analysis, do I consider the ARCH(8) model with only significant parameters, or the ARCH(9)-model with the lowest information criterion?

Thank you so much for your help!

Sonja