Allow customizing truncation max_n_steps in Hurdle Mixtures

[tjburch]

Description

Separates kwargs going to _hurdle_mixture by name. Gathers any which should redirect to Truncated (lower, upper, and max_n_steps), and otherwise assumes they should go to Mixture.

Related Issue

• Closes BUG: passing the 'max_n_steps' parameter as kwarg to HurdleNegativeBinomial distribution does not work  #7307

Reproducing their MWE after changes:

import pymc as pm
import pytensor as pt
with pm.Model():
     ad_nb = pm.HurdleNegativeBinomial('ad_nb', psi=.1, n=4000, p=1 - 5.8 * 1e-5, max_n_steps=10000)
     prior = pm.sample_prior_predictive(samples=100)
print("Done!")

Gives:

WARNING (pytensor.tensor.blas): Using NumPy C-API based implementation for BLAS functions.
Sampling: [ad_nb]
Done!

Checklist

• Checked that the pre-commit linting/style checks pass

• Included tests that prove the fix is effective or that the new feature works
  No new tests, confirmed existing tests/distributions/test_mixture.py runs without errors (python 3.12)

• Added necessary documentation (docstrings and/or example notebooks)
  Probably not warranted

• If you are a pro: each commit corresponds to a relevant logical change

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):

---

📚 Documentation preview 📚: https://pymc--7339.org.readthedocs.build/en/7339/

[ricardoV94]

upper is not supposed to be part of hurdle, so I would just do max_n_steps, which can be an explicit kwarg in the signature?

[tjburch]

Good point. Resolved in latest commit.

[tjburch]

Changing the signature led to some test fails. Looking into it...

[ricardoV94]

Perhaps a bit cleaner if you just copy the default value to the signature of the hurdle function?

[tjburch]

That's fine. Wasn't sure if we wanted to avoid doubling up on magic numbers, but if you're good with it, I am. Just made the change.

[ricardoV94]

Yes I also don't love it but seems more clean in this case where optional parameters are directed to distinct places. Otherwise we could have explicit truncated_kwargs that the user must pass as a dict.

Anyway the trick of using a Truncated should be a temporary thing, in the long term I would like to make Mixture aware of mixed types and handle it natively

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.37%. Comparing base (7d15175) to head (1087728).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7339      +/-   ##
==========================================
+ Coverage   91.25%   92.37%   +1.12%     
==========================================
  Files         102      102              
  Lines       17208    17208              
==========================================
+ Hits        15703    15896     +193     
+ Misses       1505     1312     -193     

Files	Coverage Δ
pymc/distributions/mixture.py	95.02% <100.00%> (ø)

... and 3 files with indirect coverage changes

[ricardoV94]

Thanks @tjburch !