Hey thanks for attempting to answer my rambling question. Your indexing example did help me get a better grasp on how that functions inside PyMC3. I watched another Chris Fonnesbeck video and realized where I was conceptually messed up.
I had a different goal from the tutorial, they were demonstrating how to make a prediction using the expected value, I was attempting to plot the posterior predictive sampled distribution from PyMC3 onto my original data. “theta” really is a distribution of the expected value or mean. The likelihood “y” is what I was after to complete my objective. What doubly confused me was I plotted my expected value over the original data and some of them looked good, but this was due to groups with few samples had a very wide mean estimate, while groups with more samples looked messed up due to having a more certain expected value estimate.
There is a lot to wrap my head around coming from a more traditional ML background.