LoginSign Up
← Back to all sparks
S

statsmodels

ANALYTICS
Velocity0.0

Statistical modeling and econometrics in Python

github.com/statsmodels/statsmodels ↗

statsmodels ships only what the ecosystem breaks — six releases, no new statistics.

statisticspythoncompatibilitymaintenancenumpy 2pandas 3
◆Current state
Every release in this window is a compatibility release. 0.14.2 and 0.14.3 absorbed NumPy 2, 0.14.5 fixed an import failure caused by SciPy 1.16, and 0.14.6 did the same for pandas 3.0. The only additive change across two years is Pyodide support in 0.14.4, described in its own notes as one feature and no fixes. A 0.15.0.dev0 tag exists from 2023 and has not been followed by a 0.15 release.
◆Where it's heading
The library is being kept alive rather than developed: each release answers a break introduced upstream, and the interval between them is set by the NumPy, SciPy and pandas release calendars rather than by anything statsmodels is building. Two consecutive releases whose stated purpose was restoring the ability to import the package is the sharpest available signal about maintainer bandwidth. The 0.15 line remains a dev tag with no visible progress toward a release.
◆Prediction
The next release is most likely another compatibility patch triggered by a NumPy, SciPy or pandas major, and nothing in these entries indicates 0.15 is close.

◆Recent moves

  1. 8mo ago

    0.14.6: restores importing under pandas 3.0

    A patch whose entire purpose is that statsmodels can be imported again under pandas 3.0, plus small NumPy future-proofing. The second consecutive release fixing an import break rather than shipping anything.

    View source ↗
  2. 1y ago

    0.14.5: restores importing under SciPy 1.16

    SciPy 1.16 broke the import; this release fixes it and makes some small forward-compatibility changes. No statistical functionality is added.

    View source ↗
  3. 1y ago

    0.14.4: Pyodide support

    The only additive release in this window, and the notes say so plainly — one feature, no fixes. Pyodide support puts statsmodels in the browser and in WASM runtimes, which widens where the library can be deployed even though nothing about the library itself changes.

    View source ↗
  4. 1y ago

    0.14.3: NumPy 2 environments and corrected macOS builds

    A packaging and compatibility release that makes statsmodels installable alongside NumPy 2 and recent pandas, and corrects the build procedure for macOS on both x86_64 and arm64. Unblocking builds on a whole platform is the substance here.

    View source ↗
  5. 2y ago

    0.14.2: full NumPy 2 compatibility

    Completes NumPy 2 compatibility and starts the pandas 3 future-proofing that later releases finish. It is the first entry in the compatibility-driven cadence that defines everything after it.

    View source ↗
  6. 2y ago

    Release 0.14.1

    Bug fixes accumulated since 0.14.0, with the notes stating explicitly that there are no enhancements and no API changes.

    View source ↗