Fig 8 of gelman*s paper shows results for parallelizing site computation, but the code in GitHub is only for serial computing. I want to make sure I didn’t miss something. Was the fig referring to some other code.?
The code in github performs the site updates in serial. For each such update, it may use multiple cores though. Implementing the updates to run in parallel is quite platform and problem dependent.
For the experiment of section 6.1, the code is directly provided in the github repository (the folder experiment, model m4b). For the experiment of section 6.2 (fig 8), separate code was used.