Kalman filter for PVT computation using pseudoranges and pseudorange rates

Hi everybody,
I'm a Ph.D. student involved in gps/ins integration and now i'm
working on a Kalman filter for PVT starting from pseudorange and
pseudorange rates. I have some question and I think the experience of
you all will be helpful :)

I use 8 states model (x,y,z,dt,vx,vy,vz,vt) and the model is right. I
used the model noise covariance in Brown.
The questions I have follows:
- How should I choose the pseudorange rates covariance? I compute it
from doppler estimation
- How can I get the initial point estimation?
- Which is a good starting value for the P matrix?

You can see my issues are basically on tuning :) I use the track of a
NordNav sw receiver as reference, but i don't manage to exactly follow
it :(

thanks in advance for your help
.