The R code to reproduce the results is available from GitHub Gist. Horseshoe Bend, Arizona. Photo by Madeline Pere. Although I am not an expert in Bayesian statistics, I always have an idealized version of the framework for Bayesian modeling in my mind: Allows defining data models intuitively — preferably in native R. Handles the low-level computations such as MCMC automatically. Works on both CPU and GPU seamlessly would be perfect for …