So if I understand correctly, you’re basically sampling from n Laplace distributions, but then say there is a correlation between them using a Cholesky decomposition, and shift it by \vec{\mu}? and for the actual use, I’ll use the PyMC LKJCholeskyCov class to draw samples right?
So would this method work for n > 2? The wiki on multivariate Laplace says the transformation suggested before is for a bivariate distribution only
which release of PyMC has the logp for multivariate laplace?