Via Gelman’s blog I came across this:
https://arxiv.org/abs/1806.02068
Perhaps this might interest someone here as an addition to PyMC3.
1 Like
Glancing at the Stan code it looks more like a model parameterization technique than a universal sampler. Is the algorithm able to rescale the parameter itself?
No idea! I just saw it on this morning (pre-coffee) and thought it might be useful. Sorry if not.