Skip to content

Validation: src/mors/cluster.py

This page tracks the @pytest.mark.reference_pinned tests that anchor the behaviour of mors.cluster against an analytical limit.

Test id Reference Scope
tests/test_cluster.py::test_percentile_of_uniform_ramp_matches_closed_form Analytical limit: the percentile of an evenly spaced distribution Pins the Cluster.Percentile output for an evenly spaced rotation-rate distribution to the closed form A + (B - A) * q / 100.

Re-derivation note

Cluster holds a population of stars and reports statistics of their rotation distribution. For an evenly spaced distribution running from A to B, the q-th percentile under linear interpolation is exactly

percentile(q) = A + (B - A) * q / 100

independent of the number of samples. The reference-pinned test builds a cluster whose initial rotation rates form an evenly spaced ramp, queries the percentile at a chosen q, and pins the result to that closed form. A midpoint-collapse or an off-by-one in the percentile index moves the result away from the analytic value.

Anchor type

Analytical limit. The percentile of a uniform ramp is a closed-form identity. A companion physics_invariant test checks that the percentile mapping is monotone in q and bounded within the distribution range.

Cross-references

  • src/mors/cluster.py: the Cluster class, Cluster.Percentile, and the per-quantity accessors that delegate to the member stars.
  • Using model percentiles for initial rotation describes the rotation distribution and the percentile interface.