[cdc-mitzimorris]

PopulationInfections and SubpopulationInfections now compute their renewal step themselves. Infection feedback is a pair of optional constructor arguments, infection_feedback_strength_rv and infection_feedback_pmf_rv. The infection_process argument and the InfectionProcess protocol, added in #847, are removed.

The legacy classes Infections and InfectionsWithFeedback are unchanged (except for one docstring cleanup to mathjax).

What the user writes

Without feedback, nothing changes.

With feedback, before:

builder.configure_latent(
    PopulationInfections,
    ...,
    infection_process=InfectionsWithFeedback(
        name="infections",
        infection_feedback_strength=strength_rv,
        infection_feedback_pmf=feedback_pmf_rv,
    ),
)

After:

builder.configure_latent(
    PopulationInfections,
    ...,
    infection_feedback_strength_rv=strength_rv,
    infection_feedback_pmf_rv=feedback_pmf_rv,
)

SubpopulationInfections takes the same two arguments. Its strength can be a scalar, or have shape (n_subpops,), (n_days_post_init, 1), or (n_days_post_init, n_subpops).

Renamed recorded sites

Before	After
{name}::rt_single_effective	{name}::rt_single_adjusted
{name}::rt_subpop_effective	{name}::rt_subpop_adjusted

rt_single and rt_subpop (raw Rt from the temporal processes) are unchanged. The adjusted Rt is the Rt the renewal equation uses; it equals raw Rt when feedback is off. "Adjusted" matches compute_infections_from_rt_with_feedback, which takes Rt_raw and returns Rt_adjusted.

Behavior

• Given the same random variables and seed, every recorded value is identical to passing infection_process=InfectionsWithFeedback(...) before this change. This was checked by comparing all trace sites of five configurations, three seeds each, before and after the change.
• One feedback argument without the other: ValueError at construction.
• Feedback PMF length differs from the generation interval: ValueError at sample time. Before, this was a shape error from jnp.einsum.
• Strength that does not broadcast to the shape of Rt: ValueError naming both shapes.
• infection_process passed: TypeError at construction; through PyrenewBuilder, at build().
• Extra keyword arguments to sample are unused.

Changes

• pyrenew/latent/base.py: private helpers _validate_feedback_options and _compute_renewal.
• pyrenew/latent/population_infections.py, subpopulation_infections.py: new constructor arguments, renewal through _compute_renewal, renamed sites, docstrings describing feedback.
• pyrenew/latent/infection_process.py, __init__.py: InfectionProcess removed; InfectionProcessSample kept for the legacy classes.
• Tests: tests of the removed argument removed; feedback tests rewritten for the new arguments; new tests for equivalence with InfectionsWithFeedback, zero strength, per-subpopulation strength, and each error. Factory fixtures in test/conftest.py.

Breaking change

infection_process, InfectionProcess, and the _effective sites were added in #847 and are in no tagged release.

[cdc-mitzimorris]

will close #902

[github-actions]

Thank you for your contribution @cdc-mitzimorris 🚀! Your github-pages is ready for download 👉 here 👈!
_{(The artifact expires on 2026-10-22T00:04:46Z. You can re-generate it by re-running the workflow here.)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.84%. Comparing base (0bd0853) to head (1a41c7e).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #906   +/-   ##
=======================================
  Coverage   98.84%   98.84%           
=======================================
  Files          59       58    -1     
  Lines        2242     2249    +7     
=======================================
+ Hits         2216     2223    +7     
  Misses         26       26           

Flag	Coverage Δ
unittests	98.84% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[cdc-mitzimorris]

@dylanhmorris now ready for review

[dylanhmorris]

LGTM, thanks @cdc-mitzimorris. Apologies for the slow review.