v0.2.4

✨ New Feature
• Added support for merging bins in correlated regions.
Requested by @mdgoodsell

📌 What’s Changed
• [feat] Enable merging of correlated bins — #53

🔗 Full Changelog: v0.2.3…v0.2.4

v0.2.3

This update fixes a bug during signal uncertainty insertion in correlated likelihoods.

Thanks to @mdgoodsell for pointing this out.

What's Changed

The full changelog for v0.2 can be found in documentation.

• [feat] Improve version check in #52
• [bug] Domain error during signal uncertainty insertion in #51

Full Changelog: v0.2.2...v0.2.3

v0.2.2

This release includes improvement in the $\chi^2$-test module, allowing the user to set signal yields to negative values. Previous versions did this depending on the requested region of the two-sided limits.

Additionally, NumPy v2 can now be used with Spey. In previous versions, this was constrained by the limitations of the Autograd package.

What's Changed

The full changelog for v0.2 can be found in documentation.

• [feat] Allow control over negative signal yields in #50

Full Changelog: v0.2.1...v0.2.2

v0.2.1

This release introduces specific control mechanisms for invalid covariance matrices, such as ones with infinite determinants.

What's Changed

• [bug] Issue with infinite determinant in #47

Full Changelog: v0.2.0...v0.2.1

v0.2.0

This release introduces the ability to perform two-sided statistical tests for SMEFT studies, enabling users to constrain the parameter of interest (POI) in both cases:  $\mu > \hat\mu$ and $\hat\mu > \mu$.

Next Steps:

• The transition from autograd to JAX will take place in the v0.2.x series, unlocking GPU acceleration and improved integration with machine learning frameworks.

Let me know if you’d like any further refinements!

What's Changed

• Add pre-commit in #44
• [feat] Implementation of two-sided tests with lower and upper bounds on the POI in #45

Full Changelog: v0.1.11...v0.2.0

v0.1.11

This update addresses a bug in the statistics combination module for a specific use case. When searching for the optimal value of $\hat{\mu}$ in the combined likelihood for the optimiser, the computation for $\sigma_\mu$ may crash if the Hessian matrix is non-invertible. In cases where the Hessian is non-invertible, the search for optimal $\hat{\mu}$ is halted, deferring to the optimizer to complete the estimation.

Thanks to @joes-git for pointing this out.

What's Changed

The full changelog for version 1 can be found here.

• Bugfix in stat combiner by @jackaraz in #43

Full Changelog: v0.1.10...v0.1.11

v0.1.10

This release includes compatibility updates for Python 3.12, as well as bug fixes for the exclusion limit calculator.

What's Changed

The full changelog can be found here.

• Improve compatibility for Python 3.12 by @jackaraz in #39
• Bugfix in exclusion limit calculator by @jackaraz (thanks to @joes-git) in #42

Full Changelog: v0.1.9...v0.1.10

v0.1.9

This release extends the limitations on the requirements, especially for SciPy and Autograd.

Normal and multivariate normal distributions have been added as PDFs. This update is specifically for integration with Contur. Thanks to @jonbutterworth and @joes-git.

Notes for the next release: Due to some deprecated system packages, some issues have been observed when Spey is used in Python 3.12. These issues will be fixed in the next release.

What's Changed

The changelog can be found here.

• Add Normal/MultivariateNormal PDFs by @jackaraz in #40

Full Changelog: v0.1.8...v0.1.9

v0.1.8

This version fixes numeric inconsistencies caused by signal uncertainties.

Thanks to @sabinekraml for pointing this out.

What's Changed

• Bugfix in signal uncertainties by @jackaraz in #38

Full Changelog: v0.1.7...v0.1.8

v0.1.7

This version includes bug fixes for the signal uncertainty synthesizer for default PDFs.

What's Changed

• Signal uncertainties by @jackaraz in #34
• Connection by @jackaraz in #35
• Development PR of v0.1.7 by @jackaraz in #36

Full Changelog: v0.1.6...v0.1.7