Hi Karin,
It is not clear what kind of model are you using let us assume it is a continuous PD with kin kout or emax type of model.
Simulating median curves can be easily done by asking for a table with a sequence of time and then zeroing the random effects you will get the equivalent of PRED at any time you want. To get Prediction intervals you simulate keeping the random effects as the model estimated them say 1000 times and then you compute pointwise prediction intervals.
To get confidence intervals on the median you will need to simulate with uncertainty this is currently not supported out of the box and you will need a creative approach:
- use simulation mode using the variance covariance matrix of fixed effects to simulate say 1000 times and you will get 1000 simulated medians with uncertainty then you can compute the CI using percentiles
- use the bootstrap based uncertainty so you simulate 1000 times using the 1000 different group of estimates to get 1000 medians and then you compute a CI around the median