v1.4.1

What's Changed

• Add PIT (Probability Integral Transform) plot to pted_coverage_test by @Copilot in #16

Full Changelog: v1.4.0...v1.4.1

v1.4.0

JAX arrays now supported as input!

Just pass JAX arrays into the functions and you can run PTED on GPU no problem! This works just like the previous PyTorch tensor functionality, and so makes it easy to use whatever platform you prefer.

What's Changed

• Add JAX array support to pted and pted_coverage_test by @Copilot in #14

New Contributors

• @Copilot made their first contribution in #14

Full Changelog: v1.3.2...v1.4.0

v1.3.2

What's Changed

• Add progress bar to all permutation tests by @fangyizhu in #12

New Contributors 🎉

• @fangyizhu made their first contribution in #12

Full Changelog: v1.3.1...v1.3.2

v1.3.1

What's Changed

• Add SBC reference to README for coverage test by @ConnorStoneAstro in #9
• Craete SBC histogram if requested by @ConnorStoneAstro in #10

Full Changelog: v1.3.0...v1.3.1

v1.3.0

What's Changed

• have return all also return pvalue by @ConnorStoneAstro in #7

Thanks @fangyizhu for the suggestion! See #6

Full Changelog: v1.2.0...v1.3.0

v1.2.0

What's Changed

• add KS-test comparison to README by @ConnorStoneAstro in #5
• Update p-value floor to better represent finite monte carlo p-value estimate now use (1+q)/(1+n_permute) instead of max(1,q)/n_permute
• Provide two-tail p-value option for main pted to check for overfitting

Full Changelog: v1.1.1...v1.2.0

v1.1.1

What's Changed

• check over under confident by @ConnorStoneAstro in #4

Now gives warning if coverage detects over or under confidence. Can be used to inform updates to models or sampling procedure.

Full Changelog: v1.1.0...v1.1.1

v1.1.0

What's Changed

• change coverage test to two tail p-value by @ConnorStoneAstro in #3

Breaking change

The pted_coverage_test now uses a density based two tail p-value test at the end to return a p-value. This way the test will now also return a low p-value if the posterior is underconfident rather than returning a value very close to 1. This is more in line with expectations when running the test, one would like to know specifically if the ground truth plausibly came from the posterior distribution. It is a separate question to ask if the posterior is being conservative, which could be added in the future if there is interest in such a function.

Full Changelog: v1.0.1...v1.1.0

v1.0.1

What's Changed

• docs: add type hints to more functions by @ConnorStoneAstro in #2
• Registered on Zenodo!

Full Changelog: v1.0.0...v1.0.1

v1.0.0

What's Changed

• perf: better scaling for large samples by @ConnorStoneAstro in #1
• more detailed docs

New Contributors

• @ConnorStoneAstro made their first contribution in #1

Full Changelog: v0.1.0...v1.0.0