Modeling Question

Here’s our code for the joint model.
In the running process, I have a few questions:

  1. I set parallel running to accelarate the process. However, as I have set 4 chains in MCMC, only 2 chains proceed fast and successfully while the other 2 were stuck at the warm up period. How to solve the problem? Should I set inital values for the 4 chains and how to determine the appropariate value?
  2. In specific, as you mentioned earlier, what kind of starting point is “reasonable”? Will values around the mean value of baseline sbp or dbp help?
  3. We have 210,000 observations in the longitudinal data and 12000 observations in the survival data. How much can we reduce the warmup and iteration?

Are the number of CPU cores used a multiple of four?