[R-sig-ME] starter values

[Arndt Kunick]

Thank you for the explanation about SSbiexp!
 
Primarily I fitted the model in the form below:
 
fit3 <- lmer(kosten ~ hm + I(hm^2) + np + I(np^2) + (1|j) + (1|o) + (0 + hm*np|o),  d)
 
The problem was that the quadratic polynomial function for hm and np does not fit well.
 
It seems that a non-linear model is more suitable in this case.
Is it possible to design a non-linear mixed-effect-model for two fix effects (nlin for hm and np)?
If no, is there an alternative approach?
 
I would be grateful about every advice or professional knowledge.
 
A*