fairscope 0.3.0__tar.gz

fairscope-0.3.0/.gitattributes

@@ -0,0 +1,10 @@
+# Normalize line endings: store text as LF in the repo, check out native on Windows.
+* text=auto eol=lf
+
+# Binary assets (never normalize)
+*.png binary
+*.jpg binary
+*.jpeg binary
+*.gif binary
+*.pdf binary
+*.ipynb binary

fairscope-0.3.0/.github/workflows/ci.yml

@@ -0,0 +1,40 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.9", "3.10", "3.11", "3.12"]
+
+    steps:
+      - uses: actions/checkout@v5
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v6
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install (base + dev; NLP extras intentionally excluded to keep CI light)
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e ".[dev]"
+
+      - name: Ruff (lint + import order)
+        run: ruff check .
+
+      - name: Black (format check)
+        run: black --check .
+
+      - name: Pytest + coverage (overall floor >=70%)
+        run: pytest --cov=fairscope --cov-report=term-missing --cov-fail-under=70
+
+      - name: Execute replication notebooks (nbmake)
+        run: pytest --nbmake notebooks/

fairscope-0.3.0/.gitignore

@@ -0,0 +1,55 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# Distribution / packaging
+build/
+dist/
+*.egg-info/
+.eggs/
+wheels/
+
+# Virtual environments
+.venv/
+venv/
+env/
+ENV/
+
+# Test / coverage
+.pytest_cache/
+.coverage
+.coverage.*
+htmlcov/
+coverage.xml
+.tox/
+
+# Type / lint caches
+.mypy_cache/
+.ruff_cache/
+
+# Docs build
+site/
+
+# LaTeX build artifacts (paper/)
+paper/*.aux
+paper/*.log
+paper/*.out
+paper/*.pdf
+
+# Notebooks
+.ipynb_checkpoints/
+
+# Datasets (NOT bundled; downloaded on demand by scripts/fetch_data.py)
+data/
+*.csv
+!tests/fixtures/*.csv
+
+# IDE / OS
+.idea/
+.vscode/
+.DS_Store
+Thumbs.db
+
+# Internal planning docs — never publish
+docs/internal/

fairscope-0.3.0/.pre-commit-config.yaml

@@ -0,0 +1,22 @@
+# Run `pre-commit install` once after cloning. Hooks run on every commit.
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.6.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: check-yaml
+      - id: check-toml
+      - id: check-added-large-files
+        args: ["--maxkb=512"]
+
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.5.7
+    hooks:
+      - id: ruff
+        args: ["--fix"]
+
+  - repo: https://github.com/psf/black
+    rev: 24.8.0
+    hooks:
+      - id: black

fairscope-0.3.0/CHANGELOG.md

@@ -0,0 +1,66 @@
+# Changelog
+
+All notable changes to this project are documented here. The format follows
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and the project aims to
+adhere to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+## [0.3.0] - 2026-06-27
+
+### Added
+- `fairscope.federated`: `FederatedFairnessAudit` + `FederatedReport` — a cross-node
+  (federated / multi-site) audit composing `core`: per-node DeLong AUC CIs, ECE, Brier and
+  F1; cross-node disparity (max−min AUC gap and Bonferroni-corrected pairwise unpaired
+  DeLong tests); optional per-node recalibration (temperature/isotonic) with pre/post ECE;
+  per-node AUC forest, reliability curves, and PDF export. Audits per-node predictions only
+  — no training and no privacy guarantee. Routed via `FairnessAudit(model, domain="federated", ...)`.
+- `fairscope.lending`: `LendingFairnessAudit` + `LendingReport` — a descriptive annual
+  approval-gap analysis (symmetric disparate impact per year, composing `core`) plus an
+  optional subgroup CATE via Causal Forest DML (`estimate_cate`, `econml.dml.CausalForestDML`).
+  Causal claims are conditional on the DML assumptions. `econml` is the optional
+  `fairscope[lending]` extra. Routed via `FairnessAudit(model, domain="lending", ...)`.
+- Documentation pages for the federated and lending modules, and an auto-generated API
+  reference for both.
+- Replication notebooks `notebooks/03_lending_replication.ipynb` and
+  `notebooks/04_federated_replication.ipynb` (synthetic; executed in CI via `nbmake`).
+
+## [0.2.0] - 2026-06-27
+
+### Added
+- `fairscope.nlp`: the five-axis Cross-Platform Fairness Evaluation (CPFE) protocol —
+  `CPFEProtocol` + `CPFEReport` (macro AUC/F1 and ΔAUC%, multiclass ECE, bootstrap
+  macro-AUC significance with Bonferroni correction, per-class disparate impact and
+  equalized odds), a structured per-axis `deployment_readiness()` diagnostic using P4's
+  stated reference bands (with an illustrative, configurable ΔAUC limit), and
+  gradient-saliency Jaccard attribution stability (`token_saliency` behind
+  `fairscope[nlp]`). Routed via `FairnessAudit(model, domain="nlp", ...)`.
+- Documentation site (MkDocs Material + mkdocstrings) published to GitHub Pages:
+  getting-started with a runnable example on the synthetic fixture, CPFE and healthcare
+  guides, and an auto-generated API reference. <https://rajveer-code.github.io/fairscope/>
+- Replication notebooks (`notebooks/01_healthcare_replication.ipynb`,
+  `notebooks/02_nlp_cpfe_demo.ipynb`) executed in CI via `nbmake`.
+
+## [0.1.0] - 2026-06-26
+
+### Added
+- Phase 0 public skeleton: package layout, MIT `LICENSE`, `pyproject.toml` (hatchling),
+  GitHub Actions CI (matrix py3.9–3.12), `pre-commit` config, `CITATION.cff`, README
+  skeleton, and the design overview in `docs/DESIGN.md`.
+- `fairscope.core`: DeLong AUC confidence intervals and paired/unpaired tests
+  (`delong.py`), a stratified bootstrap AUC test (`bootstrap.py`), Expected/Maximum
+  Calibration Error with reliability diagrams plus temperature-scaling and isotonic
+  recalibration (`calibration.py`), Bonferroni and Benjamini–Hochberg corrections
+  (`correction.py`), and subgroup metrics with symmetric disparate impact and equalized
+  odds difference (`metrics.py`). 100% test coverage on `core/`.
+- `fairscope.healthcare`: `HealthcareFairnessAudit` + `HealthcareReport` — a one-call
+  clinical fairness audit composing `core/` (per-subgroup DeLong CIs, ECE, Bonferroni-
+  corrected pairwise tests, Brier/F1), with report tables, an AUC forest plot,
+  reliability-curve plots, multi-page PDF export (matplotlib only), and an optional SHAP
+  summary (`fairscope[shap]`). A synthetic, seed-generated golden fixture regression-tests
+  the published direction and approximate magnitude (elderly < young AUC, gap ≈ 0.135).
+- Top-level `FairnessAudit(model, domain=...)` dispatcher (healthcare implemented).
+
+[0.3.0]: https://github.com/Rajveer-code/fairscope/releases/tag/v0.3.0
+[0.2.0]: https://github.com/Rajveer-code/fairscope/releases/tag/v0.2.0
+[0.1.0]: https://github.com/Rajveer-code/fairscope/releases/tag/v0.1.0

fairscope-0.3.0/CITATION.cff

@@ -0,0 +1,80 @@
+cff-version: 1.2.0
+message: "If you use fairscope, please cite it as below."
+title: "fairscope: subgroup-stratified, calibration-aware fairness auditing for ML models"
+abstract: >-
+  An open-source Python library that packages peer-reviewed statistical methods for
+  subgroup-stratified, calibration-aware fairness auditing — including per-subgroup
+  DeLong confidence intervals, per-subgroup Expected Calibration Error, bootstrap
+  significance testing, a subgroup-stratified interface to standard recalibration, and a
+  five-axis Cross-Platform Fairness Evaluation (CPFE) protocol.
+type: software
+authors:
+  - family-names: Pall
+    given-names: Rajveer Singh
+    orcid: "https://orcid.org/0009-0001-6762-6134"
+version: "0.3.0"
+license: MIT
+repository-code: "https://github.com/Rajveer-code/fairscope"
+keywords:
+  - fairness
+  - machine learning
+  - calibration
+  - DeLong
+  - subgroup analysis
+  - model auditing
+references:
+  # Published, final.
+  - type: conference-paper
+    title: >-
+      Comprehensive Evaluation of Machine Learning for Type 2 Diabetes Risk Prediction:
+      Large-Scale External Validation and Fairness Analysis
+    authors:
+      - family-names: Pall
+        given-names: Rajveer Singh
+      - family-names: Yadav
+        given-names: Sameer
+      - family-names: Bhalerao
+        given-names: Siddharth
+      - family-names: Sahu
+        given-names: Sourabh
+      - family-names: Ahluwalia
+        given-names: Ritu
+      - family-names: Awadhiya
+        given-names: Bhaskar
+    year: 2026
+    collection-title: "IEEE CIPHER 2026"
+    doi: "10.1109/CIPHER70417.2026.11523789"
+  # Preprints (no final venue).
+  - type: article
+    title: "IndiaFinBench"
+    authors:
+      - family-names: Pall
+        given-names: Rajveer Singh
+    year: 2026
+    url: "https://arxiv.org/abs/2604.19298"
+    identifiers:
+      - type: other
+        value: "arXiv:2604.19298 [cs.CL]"
+    notes: "Preprint."
+  - type: article
+    title: "When the Gate Stays Closed"
+    authors:
+      - family-names: Pall
+        given-names: Rajveer Singh
+    year: 2026
+    url: "https://ssrn.com/abstract=6742700"
+    identifiers:
+      - type: other
+        value: "SSRN: 6742700"
+    notes: "Preprint."
+  - type: article
+    title: "The Transaction Cost Trap"
+    authors:
+      - family-names: Pall
+        given-names: Rajveer Singh
+    year: 2026
+    url: "https://ssrn.com/abstract=6422358"
+    identifiers:
+      - type: other
+        value: "SSRN: 6422358"
+    notes: "Preprint."

fairscope-0.3.0/CONTRIBUTING.md

@@ -0,0 +1,41 @@
+# Contributing to fairscope
+
+Thanks for your interest. `fairscope` is built incrementally, one module per phase,
+with regression tests as the credibility anchor.
+
+## Development setup
+
+```bash
+git clone https://github.com/Rajveer-code/fairscope
+cd fairscope
+python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
+pip install -e ".[dev]"
+pre-commit install
+pytest
+```
+
+## Standards
+
+- **Every function:** full type hints; NumPy-style docstring with a runnable `Examples`
+  block; input validation with clear errors; no silent failures; deterministic given a
+  seed (expose `random_state` on anything stochastic).
+- **Tests are mandatory.** Where an authoritative reference value exists (e.g. DeLong
+  1988, `statsmodels` multitest), test against it — not against a paper. Where testing
+  against a paper, assert *direction + approximate magnitude* within a stated tolerance
+  and document any discrepancy. Committed data fixtures are small subsamples, labeled as
+  such in the docstring; they do not reproduce a full published run.
+- **Formatting/linting:** `black` + `ruff` (run automatically by `pre-commit`).
+- **Coverage:** enforced overall package floor ≥70% (`--cov-fail-under=70` in CI); target ≥85% on `core/`.
+
+## Honesty rules (non-negotiable)
+
+1. No invented mathematics — every method ports a published paper and cites it.
+2. No fabricated user counts, stars, downloads, or adoption claims anywhere.
+3. If a paper is ambiguous about a parameter, open an issue and ask — do not invent it.
+4. If a method does not match what a paper actually contains, name the function for what
+   it generically does and flag the mismatch.
+
+## Commits
+
+Small, logical, self-contained commits — never batch a whole phase into one commit. Each
+commit message should describe a real design decision or fix.

fairscope-0.3.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Rajveer Singh Pall
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

fairscope-0.3.0/PKG-INFO

@@ -0,0 +1,202 @@
+Metadata-Version: 2.4
+Name: fairscope
+Version: 0.3.0
+Summary: Subgroup-stratified, calibration-aware fairness auditing for ML models, grounded in peer-reviewed methods.
+Project-URL: Homepage, https://github.com/Rajveer-code/fairscope
+Project-URL: Repository, https://github.com/Rajveer-code/fairscope
+Project-URL: Issues, https://github.com/Rajveer-code/fairscope/issues
+Author: Rajveer Singh Pall
+License-Expression: MIT
+License-File: LICENSE
+Keywords: auc,calibration,delong,fairness,machine-learning,model-auditing,subgroup-analysis
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.9
+Requires-Dist: matplotlib>=3.6
+Requires-Dist: numpy>=1.23
+Requires-Dist: pandas>=1.5
+Requires-Dist: scikit-learn>=1.1
+Requires-Dist: scipy>=1.9
+Provides-Extra: all
+Requires-Dist: captum>=0.6; extra == 'all'
+Requires-Dist: econml>=0.15; extra == 'all'
+Requires-Dist: shap>=0.42; extra == 'all'
+Requires-Dist: torch>=2.0; extra == 'all'
+Requires-Dist: transformers>=4.30; extra == 'all'
+Provides-Extra: dev
+Requires-Dist: black>=24.0; extra == 'dev'
+Requires-Dist: nbmake>=1.5; extra == 'dev'
+Requires-Dist: pre-commit>=3.5; extra == 'dev'
+Requires-Dist: pytest-cov>=4.1; extra == 'dev'
+Requires-Dist: pytest>=7.4; extra == 'dev'
+Requires-Dist: ruff>=0.5; extra == 'dev'
+Requires-Dist: statsmodels>=0.14; extra == 'dev'
+Provides-Extra: docs
+Requires-Dist: mkdocs-material>=9.4; extra == 'docs'
+Requires-Dist: mkdocs>=1.5; extra == 'docs'
+Requires-Dist: mkdocstrings[python]>=0.24; extra == 'docs'
+Provides-Extra: lending
+Requires-Dist: econml>=0.15; extra == 'lending'
+Provides-Extra: nlp
+Requires-Dist: captum>=0.6; extra == 'nlp'
+Requires-Dist: torch>=2.0; extra == 'nlp'
+Requires-Dist: transformers>=4.30; extra == 'nlp'
+Provides-Extra: shap
+Requires-Dist: shap>=0.42; extra == 'shap'
+Description-Content-Type: text/markdown
+
+# fairscope
+
+[![CI](https://github.com/Rajveer-code/fairscope/actions/workflows/ci.yml/badge.svg)](https://github.com/Rajveer-code/fairscope/actions/workflows/ci.yml)
+[![Python](https://img.shields.io/badge/python-3.9%E2%80%933.12-blue.svg)](https://www.python.org/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
+[![Docs](https://img.shields.io/badge/docs-live-brightgreen.svg)](https://rajveer-code.github.io/fairscope/)
+
+**Subgroup-stratified, calibration-aware fairness auditing for machine-learning models — grounded in peer-reviewed methods.**
+
+📖 **Documentation:** <https://rajveer-code.github.io/fairscope/>
+
+`fairscope` packages statistical machinery that mainstream fairness toolkits do not expose as
+first-class, subgroup-stratified functions, and adds one novel protocol on top:
+
+- **DeLong confidence intervals** for per-subgroup AUC (fast midrank algorithm).
+- **Per-subgroup Expected/Maximum Calibration Error** with reliability diagrams.
+- **Significance testing** of subgroup performance gaps (paired/unpaired DeLong, stratified
+  bootstrap) with **Bonferroni / Benjamini–Hochberg** correction.
+- A subgroup-stratified **interface to standard recalibration** — temperature scaling
+  (Guo et al. 2017) and isotonic regression (Zadrozny & Elkan 2002), with pre/post-ECE.
+- A novel five-axis **Cross-Platform Fairness Evaluation (CPFE)** protocol.
+- One-call **domain audits**: `healthcare`, `lending`, `federated`.
+
+Only the CPFE protocol is presented as novel. Every other function ports a documented method
+and cites its source; the recalibration methods are standard, and the contribution there is the
+per-subgroup interface and pre/post-ECE reporting.
+
+> **Status — v0.3.0.** All five modules (`core`, `healthcare`, `nlp`/CPFE, `federated`,
+> `lending`) are implemented, tested, and released. 100% line coverage on the statistical core;
+> CI green across Python 3.9–3.12. See [`docs/DESIGN.md`](docs/DESIGN.md) for methods and design.
+
+## Install
+
+```bash
+pip install fairscope
+```
+
+Releases are uploaded to PyPI by the maintainer; if a version isn't available there yet,
+install from source or from the [release assets](https://github.com/Rajveer-code/fairscope/releases):
+
+```bash
+git clone https://github.com/Rajveer-code/fairscope
+cd fairscope
+pip install -e ".[dev]"
+pytest
+```
+
+The base install is light (NumPy, SciPy, scikit-learn, pandas, matplotlib). Optional extras:
+`fairscope[nlp]` (torch, transformers, captum), `fairscope[lending]` (econml),
+`fairscope[shap]`, `fairscope[docs]`.
+
+## Quickstart
+
+```python
+from fairscope.healthcare import HealthcareFairnessAudit
+
+# y_true : binary outcomes
+# y_score: the model's positive-class probabilities
+# age_group: a protected attribute, aligned row-for-row
+report = HealthcareFairnessAudit.from_scores(
+    y_true, y_score, {"age_group": age_group}
+).run()
+
+print(report.summary())     # per-subgroup AUC (DeLong CI), ECE, Brier, F1; flags the largest gap
+report.to_dataframe()       # tidy per-subgroup table
+report.plot_auc_forest()    # forest plot of per-subgroup AUC with DeLong intervals
+```
+
+Every domain is also reachable through one dispatcher,
+`FairnessAudit(model, domain=...)`, with `domain` in `{"healthcare", "nlp", "federated",
+"lending"}`. A runnable end-to-end example on a committed synthetic fixture is in the
+[getting-started guide](https://rajveer-code.github.io/fairscope/getting-started/) and in
+[`notebooks/`](notebooks/).
+
+## Modules
+
+| Module | Purpose | Status |
+|---|---|---|
+| `core/` | DeLong CI, bootstrap-AUC test, ECE/MCE + reliability, multiple-testing correction, subgroup metrics | ✅ shipped |
+| `healthcare/` | one-call clinical fairness audit + report (tables, forest & reliability plots, PDF, optional SHAP) | ✅ shipped |
+| `nlp/` | CPFE five-axis cross-platform protocol (centerpiece) + Captum attribution stability | ✅ shipped |
+| `federated/` | per-node DeLong + cross-node disparity + per-node recalibration | ✅ shipped |
+| `lending/` | annual approval-gap + subgroup CATE (Causal Forest DML) | ✅ shipped |
+
+Plotting (forest plots, reliability diagrams) currently lives in the domain reports.
+`lending`'s CATE estimation needs the optional `fairscope[lending]` extra (`econml`). The
+`federated` module audits per-node predictions only — it performs no training and provides no
+privacy guarantee.
+
+## How it differs from AIF360 / Fairlearn
+
+`fairscope` is complementary to AIF360 and Fairlearn, not a replacement: those toolkits do bias
+*mitigation*; `fairscope` does uncertainty-aware *measurement*. The table below was verified by
+inspecting the installed public APIs of **AIF360 0.6.1** and **Fairlearn 0.14.0** (checked
+2026-06; re-confirm if versions change).
+
+| Capability | AIF360 | Fairlearn | fairscope |
+|---|:---:|:---:|:---:|
+| Per-subgroup AUC confidence interval (DeLong) | no | no\* | yes |
+| Per-subgroup Expected Calibration Error | no | no | yes |
+| Subgroup significance test + multiple-comparison correction | no | no | yes |
+| Subgroup-stratified recalibration (temperature / isotonic) | partial† | no | yes |
+| Cross-platform five-axis protocol (CPFE) | no | no | yes (novel) |
+| Per-node / federated audit | no | no | yes |
+| Bias-mitigation algorithms | yes | yes | out of scope |
+
+\* Fairlearn's `MetricFrame` computes per-subgroup AUC *point estimates* (e.g.
+`roc_auc_score_group_min`), but provides no analytic (DeLong) confidence interval.
+† AIF360 ships `CalibratedEqOddsPostprocessing` (calibration-aware equalized-odds
+postprocessing), not a general per-subgroup temperature/isotonic recalibration interface.
+
+**Closest related work — `meval`** (Sutariya & Petersen, 2025,
+[arXiv:2512.17409](https://arxiv.org/abs/2512.17409)): a statistical toolbox for stratified,
+fine-grained model-performance analysis that *also* provides subgroup metric uncertainty and
+multiple-comparison corrections (with a medical-imaging focus). `fairscope` overlaps with it on
+uncertainty + significance; what `fairscope` adds is the specific DeLong AUC intervals, the
+per-subgroup calibration **and recalibration** interface, the five-axis cross-platform CPFE
+protocol, and one-call domain audits (healthcare / lending / federated).
+
+## Engineering
+
+- **Test-driven**, with regression tests anchored to authoritative reference values where they
+  exist (DeLong's worked example; `statsmodels` multiple-testing routines).
+- **100% line coverage** on the statistical core; CI runs pytest + coverage, ruff, and black
+  across Python 3.9–3.12, and executes the replication notebooks via `nbmake`.
+- Full type hints, NumPy-style docstrings with runnable examples, and explicit input validation
+  (an AUC on a single-class subgroup raises rather than returning a meaningless value).
+- Committed fixtures are **small, synthetic, and labelled as such**; no datasets or trained
+  models are bundled.
+
+## Grounded in published research
+
+`fairscope` ports methods from the author's peer-reviewed and under-review papers; it invents no
+new mathematics. Each function cites its source. Full venues and identifiers are in
+[`CITATION.cff`](CITATION.cff).
+
+- Diabetes risk prediction with external validation + fairness analysis (XGBoost, NHANES→BRFSS) — IEEE CIPHER, 2026.
+- A five-axis Cross-Platform Fairness Evaluation for mental-health NLP — under review.
+- Privacy-preserving federated learning for diabetes risk across heterogeneous nodes — under review.
+- Heterogeneous racial effects in mortgage approval (Causal Forest Double Machine Learning, HMDA) — under review.
+- Racial disparities in mortgage lending (RDD / DiD / decomposition, HMDA) — under review.
+
+## Citation
+
+If you use `fairscope`, please cite it via [`CITATION.cff`](CITATION.cff).
+
+## License
+
+[MIT](LICENSE) © 2026 Rajveer Singh Pall · ORCID [0009-0001-6762-6134](https://orcid.org/0009-0001-6762-6134)