SEO rework: meta description, comparison table, keyword-rich title#932
Open
daimon-pymclabs wants to merge 21 commits into
Open
SEO rework: meta description, comparison table, keyword-rich title#932daimon-pymclabs wants to merge 21 commits into
daimon-pymclabs wants to merge 21 commits into
Conversation
- Add HTML meta description and keywords to docs landing page via MyST frontmatter - Update H1 title to include "CausalPy", "Open-Source", "Bayesian", and "Python" for better ranking on non-branded search queries - Set html_title in Sphinx conf.py for consistent browser tab / SERP title - Add structured comparison table (CausalPy vs DoWhy vs CausalML vs pyfixest) to the docs landing page to capture "vs" and comparison search queries - Split README include with docs-end marker so comparison section is docs-only Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
PR SummaryLow Risk Overview The docs home page now sets MyST README include split via Reviewed by Cursor Bugbot for commit fed7fb8. Bugbot is set up for automated code reviews on this repo. Configure here. |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## main #932 +/- ##
==========================================
+ Coverage 95.07% 95.16% +0.09%
==========================================
Files 87 93 +6
Lines 13701 15030 +1329
Branches 812 896 +84
==========================================
+ Hits 13026 14304 +1278
- Misses 479 507 +28
- Partials 196 219 +23 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
* Add StaggeredDiD group-time effect plots Expose cohort-time ATT trajectories as a first-class plotting and summary surface so users can inspect heterogeneity hidden by event-time aggregation. Closes #937 Co-authored-by: Cursor <cursoragent@cursor.com> * Facet StaggeredDiD cohort trajectories Co-authored-by: Cursor <cursoragent@cursor.com> * Add placebo diagnostics to cohort plots Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
Use the shared scalar coercion helper for experiment score titles so pandas-like singleton scores satisfy type checking. Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Benjamin T. Vincent <inferencelab@gmail.com>
* Slim ARCHITECTURE.md and reference it from CONTRIBUTING.md Reduce ARCHITECTURE.md to a lean orientation doc and add CONTRIBUTING.md pointers so structural PRs keep it current. Fixes #843 Co-authored-by: Cursor <cursoragent@cursor.com> * Clarify that agents should not load CONTRIBUTING.md by default Point agents at AGENTS.md and ARCHITECTURE.md only; keep CONTRIBUTING.md for human contributor workflow. Co-authored-by: Cursor <cursoragent@cursor.com> * Separate AGENTS.md workflow from ARCHITECTURE.md design docs Move code-structure conventions into ARCHITECTURE.md and keep AGENTS.md focused on environment, testing, and tooling workflow. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
* ci(docs-preview): cancel when not in defined scopes of git diff * fix(docs): make RTD skip logic POSIX-sh compatible * fix(docs): make RTD skip hook fully POSIX-safe
* Add structural check scripts for exports and ARCHITECTURE inventory Introduce AST-based checkers that catch export wiring drift and stale experiment inventory tables without importing causalpy, wire them into prek/Makefile, and fix the missing SyntheticDifferenceInDifferences export in experiments/__init__.py. Closes #947 Co-authored-by: Cursor <cursoragent@cursor.com> * Check structural notes in architecture inventory Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
Bumps the actions group with 3 updates in the / directory: [j178/prek-action](https://github.com/j178/prek-action), [codecov/codecov-action](https://github.com/codecov/codecov-action) and [github/codeql-action](https://github.com/github/codeql-action). Updates `j178/prek-action` from 2.0.3 to 2.0.4 - [Release notes](https://github.com/j178/prek-action/releases) - [Commits](j178/prek-action@6ad8027...bdca6f1) Updates `codecov/codecov-action` from 6.0.0 to 6.0.1 - [Release notes](https://github.com/codecov/codecov-action/releases) - [Changelog](https://github.com/codecov/codecov-action/blob/main/CHANGELOG.md) - [Commits](codecov/codecov-action@57e3a13...e79a696) Updates `github/codeql-action` from 4.35.3 to 4.36.0 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](github/codeql-action@e46ed2c...7211b7c) --- updated-dependencies: - dependency-name: codecov/codecov-action dependency-version: 6.0.1 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: actions - dependency-name: github/codeql-action dependency-version: 4.35.5 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: actions - dependency-name: j178/prek-action dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: actions ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Warn when staggered DiD ATTs lack untreated-period support Detect calendar periods with no untreated observations, emit a clear UserWarning, and mark non-identified ATT(g, t) and ATT(e) cells with identified=False and NaN estimates instead of silently returning biased values when never-treated units are absent. Closes #938 Co-authored-by: Cursor <cursoragent@cursor.com> * Mask non-identified cells in get_plot_data_bayesian recompute path When hdi_prob differs from the stored aggregation value, pass the recomputed event-time DataFrame through _mark_non_identified_att_rows() so non-identified ATTs stay NaN. Add regression test for the non-default HDI path on a no-never-treated panel. Co-authored-by: Cursor <cursoragent@cursor.com> * Add tests for non-identified ATT coverage gaps Cover effect_summary filtering and _mark_non_identified_att_rows edge cases so Codecov patch coverage passes on the identification changes. Related: #950 (local patch coverage in prek) Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
* Refresh agent skills layout Move developer skills to .agents as the canonical shared home, retain compatibility symlinks, and add the salvaged user-facing causal investigation and dataset skills from the superseded AI config PR. Co-authored-by: Cursor <cursoragent@cursor.com> * Remove committed skill compatibility symlinks Keep .agents/skills as the only tracked developer skill location and document local symlinks as an uncommitted fallback for tools that need them. Co-authored-by: Cursor <cursoragent@cursor.com> * Remove obsolete Cursor command stubs Drop thin .cursor/commands aliases that only point to skills, including stale entries that referenced non-existent skills, now that .agents/skills is the canonical developer workflow surface. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
* Add local patch coverage gate Provide an opt-in make target and contributor guidance so patch coverage regressions can be caught locally before Codecov reports them on PRs. Closes #950 Co-authored-by: Cursor <cursoragent@cursor.com> * Prefer upstream base for patch coverage. Use the upstream tracking branch when available so fork-based contributor workflows compare local patch coverage against the PR base rather than a potentially stale fork main. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
* Redesign CausalPy method selection skill Make method routing more reliable for agents by replacing flat fast routing with ordered intake, text-based routing, explicit no-fit outcomes, and focused disambiguation references. Co-authored-by: Cursor <cursoragent@cursor.com> * Broaden CausalPy method skill trigger Make the method-selection skill easier to discover for plain-language impact questions that imply quasi-experimental causal inference. Co-authored-by: Cursor <cursoragent@cursor.com> * Add CITS method selection guidance Route comparative interrupted time-series questions to InterruptedTimeSeries with comparison series as predictors, and clarify when to prefer synthetic control. Co-authored-by: Cursor <cursoragent@cursor.com> * Tone down method skill opening prose Keep CITS covered in routing without elevating it in the main skill introduction. Co-authored-by: Cursor <cursoragent@cursor.com> * Polish method-selection skill for review on current main Rebase onto main and tighten the skill for review: trim YAML description into a triggers reference, document staggered DiD untreated-period identification support, cross-link causal-detective after a matched route, and update the skills README blurb. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
Centralise fit_intercept=False handling in BaseExperiment.__init__: clone the user's sklearn estimator (with deepcopy fallback), warn when overriding fit_intercept=True, and remove duplicated in-place mutation from DifferenceInDifferences, StaggeredDifferenceInDifferences, PanelRegression, and PiecewiseITS. Closes #664
…ort (#826) * Fix interrogate coverage by excluding .marimo and .scratch directories These directories contain exploratory/demo scripts that are not part of the main package and should not count toward docstring coverage. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Add _clone methods to BayesianBasisExpansionTimeSeries and StateSpaceTimeSeries Sensitivity checks like PlaceboInTime and OutcomeFalsification need to create fresh, unfitted copies of models. These _clone methods preserve all configuration (components, sample_kwargs, mode) while resetting fitted state. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Add random fold selection mode and fix assurance formula in PlaceboInTime Adds a "random" selection_method that randomly samples eligible placebo windows from the pre-intervention period, with constraints on minimum training fraction, minimum gap between folds, and optional period exclusion. Also fixes the assurance simulation to correctly model the alternative hypothesis as null baseline noise + expected treatment effect (theta_new + expected_effect), matching the paper's formulation. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Add OutcomeFalsification sensitivity check Introduces outcome falsification, which re-fits the experiment with alternative outcome formulas and reports their estimated effect sizes with HDI intervals. This is an informational check (no pass/fail) that lets researchers assess whether the pattern of effects across outcomes is consistent with their causal story. Inspired by the "causal detective" approach in Gallea (2026). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Add sensitivity checks article notebook Adds a developer notebook documenting the placebo-in-time and outcome falsification sensitivity check methodology with worked examples. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Update * Preparing PR * Address PR review feedback on sensitivity checks Must-fix: - pymc_models: forward user priors through BayesianBasisExpansionTimeSeries and StateSpaceTimeSeries _clone() (add priors kwarg to both __init__s) - placebo_in_time: tighten selection_method to Literal["sequential", "random"] Should-fix: - placebo_in_time: min_gap now measures observation-count distance between selected folds (tracked via each candidate's position in the sorted pre-period index), matching the test's intent - test_placebo_in_time: add targeted unit tests pinning the corrected assurance formula (true_effect = theta_new + expected_effect) - outcome_falsification: narrow the broad except to (PatsyError, FormulaException, DataException, ValueError, KeyError, RuntimeError, LinAlgError) and warnings.warn on skip so silent failures surface - outcome_falsification: add store_experiments flag so callers can drop fitted experiments and keep only the summary stats Nits: - notebook: add comment flagging the OWID URL runtime dependency - notebook: CO2 -> CO2 (with subscript) in markdown prose - test_outcome_falsification: extract shared its_context fixture - placebo_in_time: document greedy-selection failure mode in docstring and error message Tests: - add regression tests that _clone() preserves user_priors on both time-series models Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * Address PR #826 review feedback (round 2) Ben's review from 2026-04-23: PlaceboInTime (checks/placebo_in_time.py): - Add allow_overlap parameter (default False) that enforces non-overlap of pseudo-intervention windows, so random folds no longer violate the hierarchical model's exchangeability assumption by default. - Replace the "retry with a different random_seed" error path with a bounded greedy-selection retry loop (MAX_RANDOM_SELECTION_RETRIES=16) using deterministic sub-seeds derived from random_seed; failure message now names the knobs to relax (allow_overlap, min_gap, n_folds). - Warn when a pre-drawn numpy expected_effect_prior is shorter than the number of replications requested by the assurance simulation, documenting the cycling behaviour in _draw_expected_effect_samples. - Surface allow_overlap=True in __repr__ following the same "non-default only" pattern as selection_method. - Document how intervention_length falls back to data.index.max() - treatment_time when treatment_end_time is unset, shrinking the random-mode eligible window. OutcomeFalsification (checks/outcome_falsification.py): - Warn at run() when storing >= 3 fitted experiments, explaining that store_experiments=False keeps only summary statistics. - Rewrite __repr__ to hide default alpha and store_experiments flags. - Drop dead np.linalg.LinAlgError from the caught-exception tuple and the now-unused numpy import. Docs: - Wire its_place_in_time_analysis.ipynb into the ITS toctree so it stops being an orphan page. - Cross-link the notebook from sensitivity_checks.md under the "Where examples already exist" list. - Tag plot-only cells with hide-input and sampler-heavy cells with hide-output so the rendered page collapses non-essential chunks. Tests: - Pin allow_overlap default, the non-overlap invariant, the _windows_overlap helper for numeric and datetime indices, the allow_overlap opt-out, the bounded-retry reproducibility and exhaustion paths, and the expected-effect-prior cycling warning. - Pin the new OutcomeFalsification __repr__ pattern, the store_experiments memory warning at run() time, and its opt-out and below-threshold paths. - Expand the upstream-bug TODO in test_run_handles_failed_formula. Made-with: Cursor * Update its_place_in_time_analysis.ipynb * Retune sampler kwargs for ITS notebook Lower target_accept (0.94 -> 0.89), add explicit step_scale and max_treedepth=100, double chains (4 -> 8) with draws=500 and cores=16, and reduce n_changepoints_trend (18 -> 8). Re-executes the notebook so runtime/sample-count messages match the new config. * Fix 'fossilfuels' typo in ITS placebo notebook Addresses Ben's review (PR #826, 2026-04-30): "fossilfuels" -> "fossil fuels". * Fix '?.' punctuation in fog detective example Addresses Ben's review (PR #826, 2026-04-30): "?." -> "?". * Enable retina figure format in ITS placebo notebook Adds `%config InlineBackend.figure_format = 'retina'` to the Setup cell so plots render at high resolution on retina displays. Addresses Ben's review (PR #826, 2026-04-30). * Replace 'hallucinate effects' with 'false positives' phrasing Three occurrences updated to be slightly clearer per Ben's review (PR #826, 2026-04-30): the table row in 'The case', the framing question after the coal CO2 fit, and the callout-tip on Panel C. * Clarify 'estimator' vs 'model' scope in StateSpaceTimeSeries note Spells out that 'estimator' means the underlying counterfactual model (BayesianBasisExpansionTimeSeries, StateSpaceTimeSeries, ...) still inside the ITS quasi-experimental approach, and notes that ITS-level and model-specific assumptions still have to hold. Addresses Ben's review (PR #826, 2026-04-30). * Flag placebo-in-time framing in notebook introduction Adds a paragraph after the 'detective workflow' line that explicitly states the notebook showcases how placebo-in-time stress-tests the ITS identifying assumptions and helps move from 'CI excludes zero' to a defensible causal claim. Addresses Ben's review (PR #826, 2026-04-30). * Add closing sentence on placebo-in-time's role in ITS Final callout-important now closes with a sentence stating that, in ITS settings, placebo-in-time turns the unfalsifiable parallel- counterfactual assumption into something you can probe with the data you have, making causal claims more robust to skeptical readers. Addresses Ben's review (PR #826, 2026-04-30). * Explain PlaceboInTime n_folds, selection_method, min_gap in notebook Adds a short bulleted breakdown of the three knobs the next Pipeline cell uses (n_folds=3, selection_method='random', min_gap=2), explains why three folds is a reasonable default for short pre-periods, and links to the PlaceboInTime API page and the sensitivity_checks survey for the full set of options. Addresses Ben's review (PR #826, 2026-04-30). * Explain pipeline reuse in 'Did gas CO2 rise?' section Adds a short explainer right under the heading saying the next cell re-uses exactly the coal-CO2 pipeline with gas_co2 as the outcome, and what we expect the placebo result to look like if fuel substitution really happened. Addresses Ben's review (PR #826, 2026-04-30). * Expand calibration plot interpretation and decision-making notes Rewrites the 'How to read the three panels' callout to give a fuller account of what each panel encodes, with explicit decision rules for panel C, the meaning of p_cal, and the SUPPORTED threshold. Adds a 'Beyond the qualitative reading' paragraph that points to the quantitative decision tools already in PlaceboInTime (Bayesian assurance, ROPE, n_design_replications) and cross-links to the PlaceboInTime API page and the sensitivity_checks survey. Also moves the post-plot observation ('the coal CO2 effect is massive') into a new markdown cell *after* the calibration plot, so the reading order is: how-to-read -> plot -> what we actually saw. Addresses Ben's review (PR #826, 2026-04-30). * Number L2 headings 'Question 1..4' to match the case-study table Renames the four investigative L2 headings to match the four-row table in 'The case' so the link between the questions and the sections is unambiguous: Q1 -> Estimate the effect on coal CO2 Q2 -> How often does the model lie? Q3 -> Falsification --- where did the energy go? Q4 -> Did total CO2 or energy demand change? (NEW L2) Q4 was previously two ### sub-sections nested inside Q3, which mixed the trace-the-mechanism step with the boundary-check step. Promoting it makes the boundary check its own section and matches the table's row 4. The total-CO2 transition prose is rewritten to serve as the Q4 lead-in. Also demotes the standalone 'Visualising what the estimator sees when nothing happens' L2 to ### so it sits cleanly under Q2 rather than becoming a 5th top-level entry. Addresses Ben's review (PR #826, 2026-04-30). * Shrink first overview plot figsize for larger relative text Reduces the four-panel UK overview from figsize=(14, 8) to (9, 6) so axis labels and panel titles read at a more comfortable relative size (retina rendering keeps it crisp). Addresses Ben's review (PR #826, 2026-04-30). * Switch calibration plot to 3 rows and shrink figsize Reshapes plot_placebo_calibration from a 1x3 strip at (14, 4) to a 3x1 stack at (7, 9). Each panel is now wider relative to the figure (7 vs ~4.7 inches) so axis labels, legends, and the p_cal banner read at a more comfortable relative size, and the overall figure is narrower so the existing fontsize=7/8/9 settings now look right without further tuning. Addresses Ben's review (PR #826, 2026-04-30). * Refresh notebook outputs after replotting + retina config Re-runs the notebook so the embedded figures match the updated plotting code (overview at figsize=(9, 6), calibration plot in 3x1 layout) and reflect the retina figure_format. Sampler stats also refreshed; numeric results are unchanged in substance, only the small run-to-run variation expected for fresh MCMC seeds. * Fix codespell warning: 're-use' -> 'reuse' Pre-commit's codespell skips .ipynb files, but CI converts the notebook to markdown via docs/source/.codespell/notebook_to_markdown.py and runs codespell on the result, which flagged 're-use' on line 452 of the converted Q3 explainer. * Fix ruff lints in ITS placebo notebook (prek/pre-commit.ci) Four ruff errors flagged in CI by prek (the local pre-commit hook config excludes docs/, but prek lints them anyway): * I001 + F401 in cell 1 (Setup): drop unused mpatches import and normalize the import grouping (stdlib / third-party / first-party). * B905 in cell 14: add strict=True to zip(fold_results, fold_samples). The two lists are constructed in lockstep so True is the safer choice (catches mismatch as a bug rather than silently truncating). * C408 in cell 14: rewrite bbox=dict(...) as a {} literal. Verified with `ruff check docs/source/notebooks/its_place_in_time_analysis.ipynb`: 'All checks passed!'. * changes * Update outcome_falsification.py * Update placebo_in_time.py * precommit * 3.14 * Rename ITS placebo notebook and merge main Rename its_place_in_time_analysis.ipynb to its_placebo_in_time_analysis.ipynb and update notebook index and sensitivity_checks cross-links. Merge latest main to stay current with upstream. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Co-authored-by: Benjamin T. Vincent <inferencelab@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>
* Update UML Diagrams * Trigger CI (bot-opened PRs skip GITHUB_TOKEN workflow chain) Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Benjamin T. Vincent <inferencelab@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>
pre-commit-ci and other bots should not receive first-contributor welcome comments when opening autoupdate PRs. Co-authored-by: Cursor <cursoragent@cursor.com>
Supersedes pre-commit.ci autoupdate PR #869: bump ruff, pymarkdown, and mypy; keep pyproject2conda on 0.22.x until the 0.23 regression is fixed. Add blank lines around list blocks required by pymarkdown v0.9.37. Co-authored-by: Cursor <cursoragent@cursor.com>
Suppress zizmor dangerous-triggers with an inline justification, set
top-level permissions to {} and grant issues/pull-requests write at the
job level only, and hoist github.actor into env:.
Closes #888
Co-authored-by: Benjamin T. Vincent <inferencelab@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
* Broaden numpydoc-validation scope across causalpy/ (#898) Replace the ``.plot``-only exclude regex in ``[tool.numpydoc_validation]`` with a private-member exclude (``\\._``), and widen the pre-commit ``files`` pattern from ``^causalpy/experiments/.*\\.py$`` to ``^causalpy/(?!tests/|data/).*\\.py$`` so the existing ``PR01``/``PR02`` checks now apply package-wide. Document and fix every public-API ``PR01``/``PR02`` violation that surfaces under the new scope: convert remaining Sphinx ``:param:`` blocks to numpydoc ``Parameters`` sections, add missing ``**kwargs``/``*args`` rows, populate the protocol/check/maketables/skl helpers, and tidy a ``See Also`` entry in ``utils.py`` that previously broke the docstring parser. The simulated-data helpers under ``causalpy/data/`` are intentionally excluded for this round; they keep their didactic ``:param`` style. Co-authored-by: Cursor <cursoragent@cursor.com> * Enable PR04 (parameter has no type) numpydoc check (#898) Add ``PR04`` to the numpydoc-validation check set so parameter rows missing a type are caught going forward. No code changes are required: every documented parameter under the package-wide scope already declares a type. Co-authored-by: Cursor <cursoragent@cursor.com> * Enable PR07 (parameter has no description) numpydoc check (#898) Add ``PR07`` to the numpydoc-validation check set so undocumented parameter rows are caught going forward. No code changes are required: every parameter row under the package-wide scope already carries a description. Co-authored-by: Cursor <cursoragent@cursor.com> * Enable PR10 (space before colon) numpydoc check (#898) Add ``PR10`` to the numpydoc-validation check set so parameter rows that omit a space before the type colon are caught going forward. No code changes are required: existing docstrings in the broadened scope already conform. Co-authored-by: Cursor <cursoragent@cursor.com> * Enable RT03 (return value has no description) numpydoc check (#898) Add ``RT03`` to the numpydoc-validation check set so undocumented return rows are caught going forward, and flesh out the bare ``CheckResult`` return entry on the ``Check`` protocol's ``run`` method (the only violation surfaced under the package-wide scope). Co-authored-by: Cursor <cursoragent@cursor.com> * Enable SS03 (summary ends with period) numpydoc check (#898) Add ``SS03`` to the numpydoc-validation check set so first-line summaries that omit a trailing period are caught going forward, and update every docstring under the package-wide scope that previously violated the rule (mostly module-level docstrings and a handful of method summaries). Co-authored-by: Cursor <cursoragent@cursor.com> * Enable GL07 (section ordering) numpydoc check (#898) Add ``GL07`` to the numpydoc-validation check set so out-of-order docstring sections are caught going forward. Reorder ``Examples``/``Notes``/``See Also`` blocks in ``InterruptedTimeSeries``, ``PiecewiseITS``, ``PyMCModel.priors_from_data``, and the variable-selection-prior classes (the latter previously used a non-standard singular ``Example`` heading), all to match numpydoc's canonical ordering. Co-authored-by: Cursor <cursoragent@cursor.com> * Enable GL08 and replace interrogate with numpydoc-validation (#898) Add ``GL08`` (object has no docstring) to the numpydoc-validation check set so missing docstrings are caught the same way every other rule is, and retire the standalone ``interrogate`` hook now that ``numpydoc-validation`` covers the same ground (and more) for the public API. Concretely: - Drop the ``interrogate`` pre-commit hook, the ``[tool.interrogate]`` section in ``pyproject.toml``, the ``interrogate`` extras entries, the ``interrogate-badge`` workflow, the ``Makefile`` invocation, and the README badge. - Add a top-level package docstring in ``causalpy/__init__.py`` and flesh out abstract ``fit``/``build_model``/``calculate_cumulative_impact`` stubs that previously had no docstring. - Rename the nested ``plot_weights``/``make_hists`` helpers in ``InversePropensityWeighting._plot_weights_propensity`` to a leading underscore so they fall under the configured private exclude. Co-authored-by: Cursor <cursoragent@cursor.com> * Fix numpydoc violations in code merged from main (#898) After rebasing onto main, the widened numpydoc-validation scope surfaced new violations in code that landed on main after this PR was opened (SyntheticDifferenceInDifferences, the SDiD weight fitter, the basis-expansion and state-space time-series models, and the OutcomeFalsification check). Document the missing parameters and fix the SDiD module summary period. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
…add-causalimpact-comparison
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
Collaborator
Automated triageRecommendation: Why:
Review focus:
Confidence: medium |
drbenvincent
added
the
review:high
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
SEO improvements to the CausalPy docs landing page based on an audit comparing it against pymc-marketing and competitor packages:
<meta>tagshtml_titlein conf.py to match<!-- docs-end -->marker that splits the README includeTest plan
<meta name="description">and<meta name="keywords">appear in the rendered HTML<head><!-- docs-end -->comment is invisible in Markdown)🤖 Generated with Claude Code