Divergent transitions when sampling from Dirichlet distribution

[bgoodri]

@jeffeaton

Any guidance on the most efficient parameterizations would be useful.

For this case, the parameterization of the simplex that rescales gamma(alpha, 1) primitives avoids the divergences problem and the ordering arbitrariness. See

Ragged array of simplexes

Hello gurus, I would like to implement a model with a ragged array of simplexes. In response to this query about a year ago, Bob mentions (I think) that there was an intention to expose the simplex_constrain function and log determinant of the Jacobian, which would facilitate this: https://groups.google.com/forum/#!msg/stan-users/9WTI41kBf4I/sb_WAt6yBQAJ I was wondering if this was implemented? (Or perhaps I misunderstood.) Many thanks, Jeff

For the case you are probably actually referring to (i.e. with data), it is hard to tell in advance which parameterization will work best.