It’s true this tutorial is not for beginners. I didn’t write it initally, but I think its goal is just to focus on the LKJ distribution and parametrization of the MvNormal, not on the MvNormal per se. So I think its very focused scope is actually a good thing and answers a precise use-case: you’re working with MvNormal and wonder how the LKJ factor can be done with PyMC3 → boom, this is the tutorial.
This is also why I kept a bit of the old syntax: 1) it explicits what the new syntax does under the hood, to limit a “black-box” usage of pm.LKJCholeskyCov and help people understand concepts; 2) It helps people who were using the old syntax transition to the new one – as it was only introduced only in 3.9.0, we think it’s important to give people time to adapt.
Regarding tutorials dedicated to MvNormal usage, we already have good resources IMO: the Statistical Rethinking repo as you mentioned above, but also the newly revamped radon NB example. When starting out, I remember functions’ docstrings were very useful too.
That being said, we’re always happy to add good new tutorials to the catalogue, so feel free to submit one in a PR 