Repository navigation
docs(plan): design R statistical API and formula lowering over JC - #1613
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configuration
Warning Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption. 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 |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 48d02ccfa1
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| | Sample covariance from cross moments | jc::stats::sample_covariance_from_cross_power_sums | Resident covariance | | ||
| | Simple OLS and R² from cross moments | jc::stats::simple_regression_from_cross_power_sums / r_squared_from_cross_power_sums | Small model-spec subset | | ||
| | One-way ANOVA and eta² from per-group moments | jc::stats::anova_from_power_sums / eta_squared_from_power_sums | Grouped response ~ categorical predictor | | ||
| | Weak-dependence calibration | jc::jirak | Explicit noise-floor claims, not ordinary independent-sample p-values | |
There was a problem hiding this comment.
jc::jirak does not expose a weak-dependence calibration that can consume a statistical request: its only public function is prove(), which runs a fixed 5,000-sample simulation and returns a PillarResult. Consequently it cannot implement the later CalibrationId path or provide an alternative to the classical p-values, and it also contradicts the following claim that the last four entries consume moment structures. Mark this capability as missing and gate a real input-dependent calibration API before listing it among implementations to reuse.
Useful? React with 👍 / 👎.
| - CrossPowerSums / PowerSums have bounded exact integer accumulation and | ||
| checked conversions; specifically preserve the currently documented | ||
| ~2^32 rows/group bound and checked i128 centering. Overflow is refusal. |
There was a problem hiding this comment.
The existing Mask-RISC validation checks planes.n_rows > MASKED_SUM_I32_MAX_ROWS before either grouped-moment terminal runs, so a population with more than 2^32 total rows is refused even when every individual group is below this stated per-group bound. If D-JCR-1 validates only the documented group counts, such requests will pass the new contract and then fail during execution; specify the total input-plane bound in addition to JC's per-group moment constraint.
Useful? React with 👍 / 👎.
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
Purpose
Follow up the CerebroCortex-RS ACT-R/SoA integration proposal with the missing detailed R-as-JC-API plan in the repository that owns JC.
This PR is documentation only, and proposes:
pearson_from_cross_power_sums,anova_from_power_sums,simple_regression_from_cross_power_sums, reliability, effect sizes and Jirak calibration. No duplicated estimators.Relationship to existing plans
This is additive to:
.claude/plans/sql-polyfill-r2il-v1.md(procedures, not new opcodes).claude/plans/relalg-v1.md(current IR limits).claude/plans/souffle-ram-64-thought-stack-v1.md(no extra AST or foreign gather)crates/jcThe first proposed implementation PR is only D-JCR-0/1 (typed contract and parity inventory), followed by a separately gated Pearson moment probe. This PR does not implement any R bindings or claim performance improvements.
Verification
Plan:
.claude/plans/jc-r-statistical-api-v1.md