A glmm prediction problem

I'm using lmer with a generalized linear (binomial) mixed model with
nested
random effects, like:

	y ~ (1 | a / b / c)

There are no fixed effects.  After fitting the model, I would like to
make 
predictions for a new set of y values: specifically, I want to predict
BLUPs
for the random effects, and I would like to compute likelihoods for sets
of
y values under the fitted model.

I don't see a completely straightforward way of doing this since it
isn't
the usual sort of prediction problem.  Is this even a sensible thing to
do?

-- David Hinds