Repository navigation
Prepare isolated R glmGamPoi dispersion option for upstream - #187
Conversation
Add glmGamPoi 1.22.0 and eleven transitive Bioconductor/CRAN records for R 4.5.0 / Bioconductor 3.22. Keep every existing lockfile record unchanged. Include the package in dependency discovery, installer/startup diagnostics, and both CI package verification lists. This prepares the R environment for the optional NB dispersion-trend method. No statistical method or default is changed in this commit.
Keep per-taxon MASS::glm.nb maximum likelihood as the default. For the optional glmGamPoi setting, fit the shared trend across retained taxa through RCall, then use fixed-theta MASS negative-binomial Wald models. Report that trend uncertainty is not propagated and this is not the quasi-likelihood test. Refuse missing dependencies and unusable fits; withhold inference at zero dispersion, without falling back to ML. Add the config option to the API and existing frontend editor. Publish complete canonical PATCH documents atomically, validating prospective method/dispersion pairs throughout affected inherited cascades for both PATCH and DELETE. Test direct R references, invalid trends, missing-package HTTP responses, inherited updates/deletions and actual HTTP persistence. Validation so far: 499 Julia assertions passed, 13 optional R paths skipped before restoring the scientific library; 39 frontend tests, TypeScript and production build pass. Final restored-library validation follows on this working branch before upstream submission.
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
1 participant
Prepared and merged the isolated R glmGamPoi contribution into upstream/glmgampoi-r, without changing fork main. Upstream submission: JoshuaJewell/MetaManifold-WebUI#62, now ready for review after complete CI passed. Issue #37 remains open pending acceptance and merge.
The resulting commit 50f155f is Verified by GitHub, has one parent and is one commit ahead / zero behind current upstream main. Working changes were pushed in two increments before the signed preparation squash.
The change uses R glmGamPoi through RCall, leaves ML as the default, adds twelve required lockfile records with all existing pins preserved, and discloses that trend uncertainty is not propagated and this is not quasi-likelihood testing. It includes the existing config editor/API support, atomic canonical PATCH persistence, inherited PATCH/DELETE validation and regressions.
Validation: