random effects models

[Christian Ramp]

Hi!
I would like to post a few questions of how to implement a variance component analysis in the ‘usual’ model selection process.
My global model is two age class model [phi (a2 t/t) p (t)]. After fitting several constrains the models phi (a2 T/.) and phi (a2 T/T) are almost equally supported. I would like to evaluate the separate trends for both age classes using a variance component analysis.
Shall I run the random effects model separately for each age class using the global model [phi (a2 t/t)] as the starting point for each or use the random effects model for the first age class as a starting point for second age class?

Some other questions on this subject:

-running variance component analysis before or after the c-hat adjustment?

- I read one should use only general models, what about running the random effect model using the constrained model phi (a2 t/.) p (t) as starting point?

- running variance component analysis without the confounded parameters (beta terms)

- Which link function should be used? Quite a few estimates (s hat & s tilde) are on the boundaries.

Thanks in advance
Chris