I'm thinking of packaging stan and its R library rstan for Debian. Any tips? stan does hybrid monte carlo estimation. stan has a somewhat unusual architecture: the stanc compiler converts a stan program (a particular model) to C++ and the C++ compiler then builds an executable which you can run. The R package provides a wrapper around this; I'm not sure of the exact architecture but I think it builds a dynamically loadable object and loads it into R. I think both of them require the stan library (in the libstan sense, not the R sense). Ross Boylan P.S. In case you're wondering why bother when R already has a HybridMC package, stan provides automatic differentiation, compiled code, automatic tuning of hybrid monte-carlo parameters, and the "no U-turn" (NUTS) algorithm that extends the usual hybrid monte carlo approach.