opened 08:59PM - 16 Apr 22 UTC
beginner friendly
A problem with `pm.Constant` recently cropped up here: https://discourse.pymc.io…/t/typeerror-when-combining-constant-and-normal-distribution-in-function-within-deterministic-node/9219
`Constant` should only be used in special cases like Mixtures, not as a direct sampled variable. We should make this point clear in the docstrings, mentioning constant python/numpy values can be used in a PyMC models and if traces are desired `pm.Deterministic` can be used. If the goal is to show up in graphviz, `pm.ConstantData` can be used.
We can probably raise a UserWarning in `pm.Constant.__new__` to the same end.