Pymc3 produces different results than Stan/NumPyro

Specifically, I would explicitly name each argument (and generally recommend doing so):

w=pm.StudentT('w',
              nu=78.04098510742188,
              mu=-17.60015869140625,
              sigma=68.0557632446289,
              shape=(10,))

That gets me results that are pretty close to those you are getting from Stan.

1 Like