GARCH - Testing for ARCH

[nsm]

Hi guys,

Hope this is the right forum. It's primarily a question about econometrics in general, but I'm using EViews.

Would really appreciate some comments on the following: I've tested for ARCH effects in an ARMA(1,1) for different maturities in the money market. Both the F and LM statistic indicates that ARCH effects are present. I used the Residual Diagnostics -> Heterosced. Tests -> ARCH after estimating the ARMA model in EViews 7.

1) I've tried with different lag lengths, and both the F and LM have p values=0 for all lags from 1 to 9. However, only the first and second lag coefficients are significant on an individual basis (t-tests). R^2 increases as I add more lags. How do I determine the right order, or is this one of the problems with ARCH modeling?

2) My ultimate goal is to model a GARCH for differnt maturities in the money market. How do i "transfer" the ARCH results to the GARCH modeling? Let's say ARCH(2) is assumed for the ARMA model. Should I consider a GARCH(2,p) model then?

I've checked out some literature, Brooks, Gujarati etc. Great books, but they seem to be rather vague when it comes to decisions about structure and lag length (which perhaps suggests that there are no easy answers to my questions)

Thanks,

Marty