I also question your need for different variances for each treatment. Did you look at this to see if it is justified? You can compare models with and without the group effect for the random term by comparing log likelihoods, but only if you use method=laplace.
... View more
