PyMetaAnalysis 0.2.0

A pandas-first, auditable meta-analysis library for Python

PyMetaAnalysis

PyMetaAnalysis is an early-stage, pandas-first Python library for conventional study-level meta-analysis. It accepts DataFrames, NumPy arrays, and ordinary Python sequences, then returns immutable, auditable result objects containing study effects, exclusions, weights, method choices, diagnostics, provenance, and structured reports.

Early-stage: the API may change during the 0.x series. Consequential results should be reviewed against the analysis protocol and independently checked.

Install

python -m pip install PyMetaAnalysis

Install optional Matplotlib plotting support with:

python -m pip install "PyMetaAnalysis[plot]"

The distribution name is PyMetaAnalysis; the import name is meta_analyze.

Quick start

import meta_analyze as ma

result = ma.meta_analysis(
    effect=[0.12, 0.35, -0.08, 0.21],
    variance=[0.04, 0.06, 0.03, 0.05],
    study=["Trial A", "Trial B", "Trial C", "Trial D"],
    model="random",
    tau2_method="REML",
)

print(result.summary())
print(result.study_results)

DataFrame column names work directly:

result = ma.meta_analysis(
    studies,
    effect="effect",
    variance="variance",
    study="citation",
    subgroup="region",
)

Omit study= to use the DataFrame index. Supplying subgroup= returns a dedicated result containing group fits, the overall fit, and a formal test for subgroup differences.

Supported analyses

Input	Effects	Pooling/models
Effect + sampling variance or standard error	Generic	Common/random inverse variance
Two-group events + totals	OR, RR, RD	Common MH OR/RR; common/random IV
Two-group means + SDs + sizes	MD, Hedges' g	Common/random inverse variance

Random-effects inverse-variance models support REML (default), Paule-Mandel, and DerSimonian-Laird tau-squared estimators. Mean confidence intervals support the normal default plus unmodified and safeguarded Hartung-Knapp variants. Eligible random-effects fits include an HTS prediction interval.

Generic analyses accept exactly one of variance= or standard_error=. Standard errors are squared internally and the conversion is recorded in the result provenance.

Sparse binary behavior is explicit: study-level and Mantel-Haenszel continuity corrections are separate, relative-effect double-zero/double-all rows remain visible as exclusions, and RD exposes rd_zero_variance="correct" | "exclude".

Inspect and report

result.estimate
result.display_estimate
result.ci
result.tau2
result.i2
result.i2_method
result.diagnostics
result.provenance

methods_text = result.method_details()
report = result.report()
payload = report.to_dict()
json_text = report.to_json()
markdown = report.to_markdown()

OR and RR remain on the log model scale in auditable numeric attributes; display_estimate, display_ci, and display_prediction_interval provide exponentiated ratios.

Rows excluded by missing-value or sparse-data policies remain in study_results with a stable row_id, included=False, and an exclusion_reason.

Sensitivity and plots

leave_one_out = result.leave_one_out().to_dataframe()
cumulative = result.cumulative(order="publication_year").to_dataframe()

ax = result.forest(show_prediction_interval=True)
ax = result.funnel()

Plotting methods return Matplotlib axes and never call show(). Funnel plots are descriptive small-study-effect diagnostics, not proof of publication bias.

Documentation

The complete documentation is published at zhaoboding.github.io/PyMetaAnalysis.

• Installation
• Getting started
• Input data and row decisions
• Generic, binary, and continuous guides
• Choosing methods and statistical formulas
• Sensitivity analysis and plotting
• Public API, result objects, and report schema
• Validation strategy and scope/limitations
• Citation guidance
• R meta/metafor mapping

An executable end-to-end notebook uses synthetic data to demonstrate analysis, provenance, reporting, sensitivity, and plotting.

Build the complete site locally with:

python -m pip install ".[docs]"
python -m mkdocs serve

Validation status

The test suite combines hand calculations, statistical invariants, numerical edge cases, and committed R metafor reference fixtures. CI covers Python 3.10–3.13, declared dependency lower bounds, strict typing/linting, docs, and distribution builds.

This is independent cross-software validation, not a formal external statistical audit. See validation for exact coverage.

Contributing

See CONTRIBUTING.md and the full development guide. Statistical changes require formula documentation, boundary tests, and an independent comparison where available.

Security-sensitive reports should follow SECURITY.md.

License

MIT