I opened a ticket about this here: https://github.com/pymc-devs/pymc3/issues/4002
There are a number of issues all tangled together here, including:
- The use of cython and extra packages means that this notebook does not build in our environment.
- The use of cython means that this notebook is machine architecture and linux (and distro?) specific, and there’s been no attempt to make it portable.
- The use of random variables rather than (at least temporarily) constant values for “observations” is questionable.