Move slightly duplicated RVs to PyTensor

[ricardoV94]

Description

We have a few RVs implemented in PyMC with similar (or simply duplicated) forms in PyTensor. We should move those there / remove to avoid the code duplication.

• WaldRV (extra param)
• BetaClippedRV (small change in perform)
• StudentTRV (duplicated)
• WeibullBetaRV (extra param)

[yashvmanmode]

Hi @ricardoV94

I am Yash Manmode. I am a beginner in Open source Contribution. can you please give more details on this issue so that I can contribute to this issue.

thank You

[jiisa-k]

Hi @ricardoV94! If I understand correctly, these RVs have been implemented in PyMC and a similar or the same version has been implemented in PyTensor. Would you prefer moving those there and removing the implementation in PyMC or directly removing these implementations from PyMC?

[adithyalaks]

Hi, tried tackling this as part of PyData NYC '24. I think there's some uncertainty about what needs to be done here.

Taking the example of WaldRV in PyMC: We have an extra parameter alpha. But we also define a rng_fun that consumes alpha.

Then would the solution be to accept alpha as an argument here, and then simply add it to the return value here?

Or would it be to define an rng in PyTensor as is done in PyMC for WaldRV and pass it as a kwarg here rng= so that we do exactly as PyMC did?