MCMC and epsilon

[GDistiller]

Hi
My question relates to the estimation of epsilon ie the usual residual error term and to the MCMC procedure. I want to run a model to estimate survival with 3 groups (corresponding to different sites) and to set a global random effect on year that applies to all the groups as well as a site*year random effect that is unique to each site. My first query is that I am a little unsure if the unique site*year is just the usual residual variation term epsilon?

If it is then I am unsure if I need to specify the site*year unique effect in the design matrix and then pool these parameters into a hyperdistribution, or if I just leave it out of the design matrix altogether? I know from other examples that the residual variance is usually not specified in the DM and is not part of the usual model output either.

I need to estimate the size of both the global and the site*year variance components and would really appreciate any guidance on what would be the best way to set this up...

Thanks!

Greg