About Pystan Convergence Issue

[ahartikainen]

Hi,

hmc/nuts do not maximize against log_prob, but our optimize method does.

See for example 15.1 Hamiltonian Monte Carlo | Stan Reference Manual

and

The typical set and its relevance to Bayesian computation

tl;dr The typical set (at some level of coverage) is the set of parameter values for which the log density (the target function) is close to its expected value. As has been much discussed, this is not the same as the posterior mode. In a d-dimensional unit normal distribution with a high value of d, the 99% typical set looks like an annulus or sphere or doughnut, corresponding to the points whose distance from the origin is approximately \sqrt{d}. This intuition is important because it helps un…