confidence intervals of population means fitted with a nonlinear mixed-effect model (package nlme)

lancelot <renaud.lancelot at ...> writes:

Dear all,

I wonder what is the best / recommended way to compute the confidence 
intervals of population means fitted with a *nonlinear* mixed-effect 
model (function nlme, package nlme).

Parametric bootstrapping from the model (using simulate/refit);
'parametric bootstrapping' (I don't know a better term for it)
by resampling from the parameters (using sim() from the arm package);
or the delta method (there are various delta-method calculators
around, try library(sos); findFn("{delta method}")