once you have the mean and sd for each dimension, generate 5 normal(m,s) observations for each dimension.

specify your model in Stan to precisely fit the way you generated the data... I suspect Stan will struggle, it will need tiny step sizes because some variables will be sd of 0.002 or whatever, but some variables will be WAY far away from 0 and some variables will have very large sd...

]]>I'm interested in collecting a few of these examples and trying out some alternatives.

]]>