Take my suggestion as just that: a suggestion. If you aren’t going to be more confident in a numpy-generated data set than you would be in a PyMC-generated data set, none of what I said above really applies. But given that you said you had less experience with PyMC, you might want to start somewhere you feel a bit more confidence and then use that data set (that you are more confident in) to test how your PyMC model is doing.
2 Likes