Ha! Fully appreciate that Bob, and I will dig deeper over the summer. Here, I stayed in the conjugate world for replicability of my ‘trusted’ Gibbs results.
On a related note, I’m doing some portfolio choice models where Gibbs makes the application scalable to larger choice sets by augmenting the utilities of the components of the portfolios instead of working with the 2^J potential combinations! There’s still life in the old beast 