Two-stage Bayesian regression enforcing a fixed distribution (not Just Hierarchical regression)

[junpenglao]

@ricardoV94

Otherwise, a RV in a Aesara graph would evaluate to a new value everytime it’s called.

I guess what I dont understand is that why it cannot be a Stochastic node that evaluate to new value without a Step method. In another word it behaves stochastically in forward graph evaluation but it is not a RV in the logp graph