--- id: empirical-calibration version: "0.12" license: Apache 2.0 license_treatment: permissive maintenance: abandoned --- # empirical-calibration — Package for empirical calibration License: permissive · Maintenance: abandoned · Downloads: 319.8K/mo ## What it is and what it does Empirical calibration is a statistical weighting method that corrects bias in survey samples and observational data by reweighting observations to match target population characteristics. The package solves the calibration problem as a convex optimization task in dual form, supporting multiple objectives, weight clipping, and inexact calibration. It depends on numpy, scipy, scikit-learn, pandas, patsy, absl-py, and six to handle numerical computation and data manipulation. The package is designed for survey sampling (where samples are unrepresentative) and causal inference (where treatment assignment correlates with covariates). However, the repository is archived and maintenance is abandoned as of May 2024, meaning no bug fixes, security updates, or compatibility work with newer Python or dependency versions will occur. Use it for: - Reweight survey respondents to match known population distributions when the sample is biased or non-random. - Balance covariate distributions between treatment and control groups in observational causal studies. - Correct for selection bias in administrative or real-world datasets where sampling was not random. - Compute inverse-probability weights for observational studies with complex covariate imbalance. - Apply weight clipping to avoid extreme weights that can destabilize downstream analysis. ## Worth the install? AI-flagged interpretation of the facts above — verify before relying. Computes empirical calibration weights to correct bias in survey samples and observational data through convex optimization, supporting weight clipping and inexact calibration. No. The package is abandoned (archived repository, no commits since May 2024) and has high install friction with seven dependencies. While empirical calibration is a legitimate statistical technique, the lack of maintenance means no support for modern Python or dependency versions, no bug fixes, and no security updates. Use only if you have a frozen environment and cannot migrate to actively maintained alternatives. ## Install pip install empirical-calibration uv add empirical-calibration poetry add empirical-calibration ## Installing empirical-calibration Before you install: High install friction due to 7 runtime dependencies (numpy, scipy, scikit-learn, pandas, patsy, absl-py, six). Repository is archived and maintenance is abandoned as of 2024-05-05, with no active development or support. License in practice: Apache 2.0 permissive license allows commercial and private use with minimal restrictions, though the abandoned status means no ongoing legal or security updates. Quickstart: pip install empirical-calibration import empirical_calibration as ec # Use ec to compute calibration weights for your biased sample data Requires Python 3.6 or later; package is abandoned and no longer maintained. Verify before relying: - Current compatibility with Python versions beyond 3.7 (classifiers list 3.6 and 3.7 only; package is archived) - Whether the package works reliably with modern versions of its dependencies (numpy, scipy, scikit-learn, pandas) - Performance or robustness claims relative to existing software mentioned in the description ## Package facts - License: Apache 2.0 (permissive) - Python support: unspecified - Install friction: high - Maintenance: abandoned - Downloads: 319.8K/month (top 15,000 on PyPI) - Known vulnerabilities: none known ## Tags survey sampling bias correction, empirical calibration weights, observational data weighting, causal inference covariate balance, sample bias adjustment, convex optimization weighting, survey calibration python, survey-sampling, causal-inference, bias-correction [View on SkillFed](https://skillfed.io/packages/empirical-calibration) · [View on PyPI](https://pypi.org/project/empirical-calibration/)