InferenceData from prior predictions without constant data

[TimOliverMaier]

constant_data missing after pm.sample_prior_predictive()

In some use-cases one might want to sample and plot prior predictions before sampling the posterior. For
this purpose, it would be nice to have "constant_data" group in the az.InferenceData object, that
pm.sample_prior_predictive() returns.

Please provide a minimal, self-contained, and reproducible example.

import pymc as pm
import numpy as np

x = np.arange(-10,10)
y = 2*x+1.4+np.random.normal(0,0.3,x.size)

with pm.Model() as model:
    x = pm.MutableData("x",x)
    y = pm.MutableData("y",y)
    a = pm.Normal("a",mu=1,sigma=2)
    b = pm.Normal("b",mu=0,sigma=2)

    mu = pm.Deterministic("mu",var=a*x+b)
    e = pm.Exponential("e",lam=10)

    obs = pm.Normal("obs",mu=mu,sigma=e,observed=y)

with model:
    prior_pred = pm.sample_prior_predictive(model=model)

In the example above prior_pred has following groups:
+prior
+prior_predictive
+observed_data

Versions and main components

• PyMC/PyMC3 Version: d322f7 (current main branch v4)
• Aesara/Theano Version: e5ebf2 (current main branch)
• Python Version: 3.10.4
• Operating system: Ubuntu 20.4
• How did you install PyMC/PyMC3: (conda/pip) pip -e