PPC for data with known measurement uncertainty

[junpenglao]

@jwwang

As I found in pymc3 documents, this is probably because the shape of shared variable cannot be updated when the shared variable involves the shape of another variable (Y_max).

Yes you are right, however i dont see why you need to specify the shape of Y_max, maybe try:

Y_max = pm.Uniform('Y_max', lower=0, upper = X_shared * a + b)