varguid 0.1.8__tar.gz

varguid-0.1.8/.github/dependabot.yml

@@ -0,0 +1,10 @@
+version: 2
+updates:
+  - package-ecosystem: github-actions
+    directory: /
+    schedule:
+      interval: monthly
+    groups:
+      actions:
+        patterns:
+          - "*"

varguid-0.1.8/.github/workflows/ci.yml

@@ -0,0 +1,76 @@
+name: CI
+
+on:
+  push:
+  pull_request:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ci-${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  tests:
+    name: Python ${{ matrix.python-version }}
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.12", "3.13", "3.14"]
+    steps:
+      - name: Check out source
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+      - name: Install package and test dependencies
+        run: python -m pip install --upgrade pip && python -m pip install -e ".[dev]"
+      - name: Run tests
+        run: python -m pytest
+
+  quality:
+    name: Lint and type-check
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out source
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.13"
+          cache: pip
+      - name: Install package and development dependencies
+        run: python -m pip install --upgrade pip && python -m pip install -e ".[dev]"
+      - name: Ruff
+        run: ruff check src tests && ruff format --check src tests
+      - name: Mypy
+        run: mypy src tests
+      - name: Check source manifest
+        run: check-manifest
+
+  build:
+    name: Build distributions
+    runs-on: ubuntu-latest
+    needs: [tests, quality]
+    steps:
+      - name: Check out source
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.13"
+          cache: pip
+      - name: Install build tools
+        run: python -m pip install --upgrade pip build twine
+      - name: Build and validate
+        run: python -m build && python -m twine check dist/*
+      - name: Upload distributions
+        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7.0.0
+        with:
+          name: python-package-distributions
+          path: dist/
+          if-no-files-found: error

varguid-0.1.8/.github/workflows/python-publish.yml

@@ -0,0 +1,70 @@
+name: Upload Python Package
+
+on:
+  workflow_dispatch:
+  release:
+    types: [published]
+
+permissions:
+  contents: read
+
+jobs:
+  release-build:
+    name: Build and validate release distributions
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Check out source
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install build and test tools
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install build twine pytest
+          python -m pip install -e ".[dev]"
+
+      - name: Run tests
+        run: python -m pytest -q
+
+      - name: Build release distributions
+        run: python -m build
+
+      - name: Check release distributions
+        run: python -m twine check dist/*
+
+      - name: Upload distributions
+        uses: actions/upload-artifact@v4
+        with:
+          name: release-dists
+          path: dist/
+
+  pypi-publish:
+    name: Publish release distributions to PyPI
+    runs-on: ubuntu-latest
+    needs:
+      - release-build
+
+    permissions:
+      id-token: write
+      contents: read
+
+    environment:
+      name: pypi
+      url: https://pypi.org/project/varguid/
+
+    steps:
+      - name: Retrieve release distributions
+        uses: actions/download-artifact@v4
+        with:
+          name: release-dists
+          path: dist/
+
+      - name: Publish release distributions to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: dist/

varguid-0.1.8/.gitignore

@@ -0,0 +1,26 @@
+# Python bytecode and caches
+__pycache__/
+*.py[cod]
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+
+# Environments and local tooling
+.venv/
+venv/
+.nox/
+
+# Coverage
+.coverage
+.coverage.*
+htmlcov/
+
+# Build products and generated metadata
+build/
+dist/
+*.egg-info/
+
+# Editors and operating systems
+.DS_Store
+.idea/
+.vscode/

varguid-0.1.8/.pre-commit-config.yaml

@@ -0,0 +1,15 @@
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.15.18
+    hooks:
+      - id: ruff-check
+        args: [--fix]
+      - id: ruff-format
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+      - id: end-of-file-fixer
+      - id: trailing-whitespace
+      - id: check-yaml
+      - id: check-toml
+      - id: check-added-large-files

varguid-0.1.8/CHANGELOG.md

@@ -0,0 +1,61 @@
+# Changelog
+
+All notable changes to this project are documented here.
+
+## 0.1.8 - 2026-06-22
+
+### Fixed
+
+- Preserved DataFrame feature identity during prediction by reordering known
+  columns and rejecting missing or unexpected columns.
+- Honored no-intercept Patsy formulas such as `y ~ 0 + x1 + x2`.
+- Made formula fitting and prediction reject missing values instead of silently
+  dropping affected rows through Patsy.
+- Matched the attached R non-lasso update order after the convergence threshold
+  is reached and removed variance-prediction clipping.
+- Added a safe fallback for an exactly zero fitted variance model.
+- Standardized lasso features and added shuffled reproducible CV folds.
+- Added the Scientific Python `rng` interface for functional lasso fits while
+  retaining `random_state` as a deprecated compatibility alias.
+- Corrected scikit-learn estimator behavior: mixin order, `NotFittedError`,
+  `n_features_in_`, feature-name checks, and fitted attributes.
+- Restored keyword compatibility for `predict(object=...)` and `prd(object=...)`.
+- Corrected the nox build session so Twine receives expanded artifact paths.
+- Preserved the historical `generate_cobra2d(n_features < 10)` behavior by
+  expanding the generated design to ten predictors.
+
+### Added
+
+- Independent regression tests for the attached R non-lasso algorithm.
+- Full scikit-learn estimator checks.
+- README example, formula, schema-validation, lasso scale-invariance, RNG, and
+  exact-fit regression tests.
+- A Trusted Publishing release workflow with least-privilege permissions and
+  immutable GitHub Action commit pins.
+- A separate SLSA build-provenance attestation and verification job, plus
+  Dependabot tracking for pinned GitHub Actions.
+- A PEP 561 `py.typed` marker and typed-package classifier.
+- Release, contribution, fix-log, and validation documentation.
+
+### Changed
+
+- Adopted the publication title `{varguid}: Variance-Guided Regression Improving Upon OLS and ANOVA for Python`.
+- Updated canonical repository links to `zionwzz/varguid-python` and added machine-readable and BibTeX citation metadata.
+- Updated a test annotation for compatibility with mypy 2.1.0; runtime behavior is unchanged.
+- Raised the supported Python floor to 3.12 and refreshed core dependency floors
+  in line with the project's Scientific Python support policy.
+- CI now targets Python 3.12, 3.13, and 3.14 and separates tests, quality checks,
+  distribution building, and publishing.
+- Clean submitted source archives no longer contain generated caches, bytecode,
+  egg-info directories, or previously built artifacts. Standard distribution metadata
+  generated during an sdist build is retained as required by the build backend.
+
+### Compatibility note
+
+Sparse fits use scikit-learn's lasso solver. The implementation now mirrors the
+R workflow more closely through standardization and CV behavior, but it is not
+bit-for-bit equivalent to `glmnet`.
+
+## 0.1.7
+
+- Previous Python source release supplied for audit.

varguid-0.1.8/CITATION.bib

@@ -0,0 +1,20 @@
+@software{wang_lu_2026_varguid_python,
+  author    = {Wang, Zihao and Lu, Min},
+  title     = {{varguid}: Variance-Guided Regression Improving Upon OLS and ANOVA for Python},
+  version   = {0.1.8},
+  year      = {2026},
+  publisher = {Zenodo},
+  doi       = {10.5281/zenodo.20816141},
+  url       = {https://doi.org/10.5281/zenodo.20816141}
+}
+
+@article{liu_lu_2026_varguid,
+  author  = {Liu, Sibei and Lu, Min},
+  title   = {Variance-Guided Regression for Heteroscedastic Data With a Grouping-Based Extension for Nonlinear Prediction},
+  journal = {Statistics in Medicine},
+  volume  = {45},
+  number  = {13-14},
+  pages   = {e70632},
+  year    = {2026},
+  doi     = {10.1002/sim.70632}
+}

varguid-0.1.8/CITATION.cff

@@ -0,0 +1,40 @@
+cff-version: 1.2.0
+message: "If you use varguid, please cite both this software and the method paper listed in references."
+type: software
+title: "{varguid}: Variance-Guided Regression Improving Upon OLS and ANOVA for Python"
+version: 0.1.8
+doi: "10.5281/zenodo.20816141"
+date-released: "2026-06-23"
+authors:
+  - family-names: Wang
+    given-names: Zihao
+  - family-names: Lu
+    given-names: Min
+abstract: >-
+  A Python implementation of variance-guided regression for heteroscedastic
+  data, including baseline and variance-guided linear prediction, optional
+  lasso regularization, formula support, coefficient inference for non-lasso
+  fits, and a scikit-learn-compatible estimator.
+repository-code: "https://github.com/zionwzz/varguid-python"
+url: "https://github.com/zionwzz/varguid-python"
+license: GPL-2.0-or-later
+keywords:
+  - regression
+  - heteroscedasticity
+  - weighted least squares
+  - mean-variance modeling
+  - lasso
+references:
+  - type: article
+    authors:
+      - family-names: Liu
+        given-names: Sibei
+      - family-names: Lu
+        given-names: Min
+    title: "Variance-Guided Regression for Heteroscedastic Data With a Grouping-Based Extension for Nonlinear Prediction"
+    journal: "Statistics in Medicine"
+    volume: 45
+    issue: "13-14"
+    start: e70632
+    year: 2026
+    doi: 10.1002/sim.70632

varguid-0.1.8/CONTRIBUTING.md

@@ -0,0 +1,53 @@
+# Contributing
+
+## Development environment
+
+Create and activate a Python 3.12+ virtual environment, then install the package
+and development dependencies:
+
+```bash
+python -m pip install --upgrade pip
+python -m pip install -e ".[dev]"
+```
+
+## Required checks
+
+Run these commands from the repository root before submitting a change:
+
+```bash
+python -m pytest
+ruff check src tests
+ruff format --check src tests
+mypy src tests
+check-manifest
+python -m build
+python -m twine check dist/*
+```
+
+The same checks are grouped into nox sessions:
+
+```bash
+nox -s tests lint build
+```
+
+Install the local hooks once with `pre-commit install`, then run
+`pre-commit run --all-files` before committing.
+
+## Regression expectations
+
+Changes to the fitting algorithm should include tests for:
+
+- agreement with the update order in the attached R `R/irls.R` implementation;
+- DataFrame feature-name safety;
+- formula intercept behavior;
+- lasso reproducibility and scaling behavior, when applicable; and
+- the scikit-learn estimator contract.
+
+Sparse fits use scikit-learn rather than R's `glmnet`; document any intentional
+behavioral difference instead of asserting bit-for-bit cross-language parity.
+
+## Releases
+
+Update both `pyproject.toml` and `src/varguid/__init__.py`, add an entry to
+`CHANGELOG.md`, and follow `RELEASING.md`. Do not upload from a developer
+machine when the Trusted Publishing workflow is available.

varguid-0.1.8/FIX_LOG_0.1.8.md

@@ -0,0 +1,124 @@
+# varguid 0.1.8 fix log
+
+Audit date: 2026-06-22
+
+Inputs reviewed:
+
+- Python source archive `varguid_py_0.1.7_source.zip`.
+- R source package `varGuid_0.1.5.tar.gz`, especially `R/irls.R` and `R/add.R`.
+
+## Corrections
+
+### 1. DataFrame prediction schema
+
+**Problem:** prediction used positional columns. Reordering the same DataFrame
+columns silently changed predictions.
+
+**Correction:** fitted DataFrame names are retained. Prediction reorders a
+complete matching schema and raises a clear error for missing, additional, or
+ambiguous string-equivalent names.
+
+**Regression coverage:** `tests/test_validation.py`.
+
+### 2. Formula intercept handling
+
+**Problem:** `lmv_formula("y ~ 0 + x1 + x2", ...)` still inserted an intercept.
+
+**Correction:** the Patsy design matrix determines `fit_intercept`; no-intercept
+formulas now remain no-intercept during fitting, summaries, and prediction.
+Design information is retained for transformed and categorical predictors.
+
+**Regression coverage:** `tests/test_basic.py` and formula prediction tests.
+
+Formula fitting and prediction now use Patsy's `NA_action="raise"`, preventing
+missing values from silently reducing the number of fitted or predicted rows.
+
+### 3. Non-lasso R algorithm parity
+
+**Problem:** the Python implementation clipped fitted variance values and used a
+different model update order once the coefficient-change threshold was met.
+
+**Correction:** variance fits are no longer clipped. Candidate mean models are
+assigned before the convergence branch, matching the attached `R/irls.R`
+control flow. The last committed model remains the returned variance-guided fit.
+An exactly zero variance fit receives a finite uniform-weight fallback.
+
+**Regression coverage:** `tests/test_r_compatibility.py` contains an independent
+translation of the attached R non-lasso procedure and exercises the
+post-convergence branch.
+
+### 4. Lasso scaling and randomness
+
+**Problem:** predictors were sent to `LassoCV` without glmnet-like
+standardization, and `random_state` had no practical effect under the previous
+solver configuration.
+
+**Correction:** mean and variance predictors are standardized using the
+applicable sample weights, coefficients are mapped back to original units, and
+cross-validation uses shuffled `KFold` splits generated from a deterministic
+seed stream. The functional API now accepts the Scientific Python `rng`
+keyword, normalizes it with `numpy.random.default_rng`, and retains
+`random_state` as a deprecated alias. The scikit-learn estimator keeps its
+conventional `random_state` parameter. The R default of 3 folds for at most 80
+rows and 10 otherwise is retained when `cv_folds=10`.
+
+**Limitation:** scikit-learn and `glmnet` use different solvers and lambda-grid
+implementations. Sparse coefficients are reproducible within Python but are not
+claimed to be bit-for-bit identical to R.
+
+**Regression coverage:** lasso smoke, reproducibility, and unit-invariance tests.
+
+### 5. Scikit-learn estimator contract
+
+**Problem:** the wrapper had the wrong mixin order, raised the wrong pre-fit
+exception, omitted `n_features_in_`, and did not fully validate prediction
+features.
+
+**Correction:** `VarGuidRegressor` now inherits `RegressorMixin` before
+`BaseEstimator`, uses scikit-learn validation helpers, records standard fitted
+attributes, and raises `NotFittedError` before fit.
+
+**Regression coverage:** the full `sklearn.utils.estimator_checks.check_estimator`
+suite plus focused wrapper tests.
+
+### 6. Packaging, CI, and release security
+
+The Patsy lower bound was raised from `1.0` to `1.0.1` because Patsy 1.0.0 is a
+yanked release.
+
+**Problem:** the supplied archive included old `dist/`, cache folders, bytecode,
+and egg-info; configured lint/type checks failed; release instructions relied on
+manual token-based uploads.
+
+**Correction:** lint and type issues were resolved, the nox build glob was
+corrected, source packaging was cleaned, and CI now tests supported Python
+versions. Release publishing uses a separate least-privilege OIDC job through a
+protected `pypi` environment. Third-party actions are pinned to full immutable
+commit SHAs and tracked by Dependabot. A separate credential-limited job
+generates and verifies SLSA build-provenance attestations before the PyPI
+publishing job can run. The wheel includes a PEP 561 `py.typed` marker.
+
+### 7. Synthetic-data compatibility
+
+**Problem:** an early validation change rejected `n_features < 10`, whereas the
+0.1.7 helper and attached R generator expand such requests to ten predictors.
+
+**Correction:** positive feature counts below ten are again expanded to ten.
+Input validation, the preferred `rng` interface, and valid negative
+equicorrelation support are retained.
+
+**Regression coverage:** `tests/test_validation.py`.
+
+## Validation scope
+
+The final release was validated with pytest, Ruff lint and format checks, mypy,
+package build, Twine metadata checks, archive-content inspection, wheel
+installation, source-distribution installation, and installed-package execution
+of the documented examples. Exact commands and outputs are recorded in
+`VALIDATION_LOG_0.1.8.txt`.
+
+The available local interpreter was CPython 3.13.5. Python 3.12 and 3.14 are
+configured in CI but were not available for local execution in this audit.
+An R interpreter was not available locally; non-lasso parity was checked against
+an independent Python translation of the attached R source rather than by
+executing R itself.