ARMA Model in State Space with Constant Term

[rzima]

First post... My master's degree may be hinging on some of you guys! =)

I'm trying to model seasonally adjusted and non-seasonally adjusted data using an ARMA(3,3) model with constant term. The output for the standard LS estimate is very nice. Here are the two questions I have:

#1. Should the constant term simply be the sample mean? In the cases I looked at, it wasn't, however they were very close to one another.
#2. I can take an ARMA model into state space form, but is there a way to account for the additional constant term in the state space that I implement in the ARMA(3,3) model.

That's all I have. The goal is to take the state-space form and apply a kalman filter that will provide a much nicer forecast on the non-seasonally adjusted data. The goal is to compare the method of fitting an ARMA(3,3) model to seasonally adjusted data vs. running a kalman filter on non-seasonally adjusted data.