PCNtoolkit version 1.4.0 is out! 🎉

Announcements
1

PCNtoolkit version 1.4.0 is out! :tada:

You can update by running pip install --upgrade pcntoolkit

Some nice new features:

  • Longitudinal normative modelling (with the help of GSoC 2026!). You can now
    use the velocity models from Bayer et al., 2026 (preprint) and the z-diff score from Bučková et al., 2025.
  • A new website design with improved tutorials and more educational content
    to help you with normative modelling. You can find them here:
    https://pcntoolkit.readthedocs.io
  • We now support Python 3.12 and 3.13 (Python 3.11 is dropped).

Changelog: Release Version 1.4.0 · predictive-clinical-neuroscience/PCNtoolkit · GitHub

1 Like
2

A small reproducibility suggestion from the current dev docs: CorrelationMatrix.load() is documented as restoring the CSV values but leaving bandwidth, n_subjects and estimated_range unset. Could save/load keep those in a sidecar too? If someone passes a matrix on, it would help the next user distinguish an estimated correlation from an out-of-range convention without reconstructing where the CSV came from.