Skip to content

Ecosystem update — September 2026

The related gpbiometricspy package now includes a fully exact-main-certified crossed participant–item Gaussian hierarchical location–scale model with one location random slope for each crossed factor.

This is an ecosystem-level addition, not a new gp3sequencespy sequence model. gp3sequencespy remains focused on ordered categorical sequences, scanpaths, motifs, transition structure, distances, clustering, randomization and latent-state workflows. The gpbiometricspy method instead models continuous repeated outcomes with crossed participant/item heterogeneity in conditional location and residual scale.

Certification record

Certified gpbiometricspy PR #129 is pinned to merge SHA d078e0366ace49c3ebeb2f6800bad6394d70631e:

  • 14/14 exact-main push workflow families green;
  • 12/12 Ubuntu/macOS/Windows × Python 3.11–3.14 test lanes green;
  • 782/782 tests;
  • 14,015/14,015 statements;
  • 6,757/6,776 raw branches = 99.7196%;
  • 19 unchanged audited structural arcs with 0 unexpected, 0 stale and 0 unaudited debt;
  • frozen gpbiometrics 2.0.0 parity unchanged at 406/406.

Read the crossed participant–item random-slope guide

Open gpbiometricspy PR #129

A study may therefore use sequence methods for ordered process structure and, where scientifically justified, a separate location–scale model for continuous repeated outcomes without conflating the two inferential targets.