Dear Pierre,
Thank you very much for the reply. The suggestions are helpful, these unlikely, very large estimates vanish after the categorical trait was placed in second and the prior for the residual variance was fixed. Unfortunately the error messages ? Error in cholesterols.default (W): the leading minor of order 3 is not positive definite? were still shown when I checked the convergence for random effects with german.diag(), I wonder if it indicates that more iterations are required.
Before I proceed for a longer run, I was wondering would you (or any other list member) mind further explaining for me why my priors are strong?