Logccdf stability improvements

[maresb]

Extend _logccdf usage to Truncated and transformed discrete distributions, improve stability

Description

This PR extends the _logccdf dispatcher implementation to further improve numerical stability for log survival function computations.

Following a thorough audit of the codebase for log(1 - CDF) patterns, the following additional improvements have been made:

• Truncated distributions: Updated TruncatedRV to use the stable _logccdf for its lower bound calculation, as suggested by @ricardoV94.
• Transformed discrete distributions: The measurable_transform_logcdf in pymc/logprob/transforms.py now correctly uses _logccdf_helper for discrete variables, replacing a numerically unstable pt.log1mexp(_logcdf_helper(...)) call.
• General stability in transforms: Replaced a pt.log(pt.exp(a) - pt.exp(b)) pattern with logdiffexp(a, b) in measurable_transform_logcdf to prevent nan results at support boundaries (e.g., logdiffexp(-inf, -inf) now correctly returns -inf).

These changes ensure more robust log-probability calculations across a wider range of distributions and transformations, particularly at extreme tail values.

Related Issue

• Related to Add _logccdf dispatcher for numerically stable log survival function in censored distributions pymc-devs/pymc#7996 (This PR is an update to Add _logccdf dispatcher for numerically stable log survival function in censored distributions pymc-devs/pymc#7996)
• Closes #

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works (existing test_transforms.py tests pass)
• Added necessary documentation (docstrings and/or example notebooks)
• 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--2.org.readthedocs.build/en/2/

[cursor]

Cursor Agent can help with this pull request. Just @cursor in comments and I'll start working on changes in this branch.
_{Learn more about Cursor Agents}