PyAutoStat

PyAutoStat analyzes pandas DataFrames and returns structured statistical results, data quality findings, and reports. It supports exploratory analysis and independent group comparisons. The package is in alpha; review assumptions and results before using them in research or decisions.

Features

• Explore data: descriptive statistics, missing values, duplicates, distributions, histograms, and advisory column type and role detection.
• Check assumptions: Shapiro-Wilk, D'Agostino-Pearson, and Anderson-Darling normality results; IQR, Z-score, and MAD outlier summaries.
• Study relationships: Pearson, Spearman, and Kendall correlations, with p-values for Pearson pairs.
• Compare independent groups: automatic or explicit t-test, Mann-Whitney U, one-way ANOVA, and Kruskal-Wallis; assumption checks, effect sizes, and confidence intervals.
• Test categorical association: Pearson chi-square, Cramér's V, and Cohen's h for a two-by-two table with a named success outcome.
• Share results: severity-rated insights and dictionary, JSON, CSV, static HTML, or optional Plotly HTML reports.

Installation

Requires Python 3.10 or newer:

python -m pip install pyautostat

For interactive HTML reports, install the optional Plotly dependency:

python -m pip install "pyautostat[report]"

Quick start

This example runs without an input file:

import pandas as pd

from pyautostat import InsightEngine, ReportGenerator, StatisticalAnalyzer

df = pd.DataFrame(
    {
        "group": ["control"] * 8 + ["treatment"] * 8,
        "outcome": [4, 5, 5, 6, 4, 5, 6, 5, 7, 8, 7, 9, 8, 7, 9, 8],
    }
)

analyzer = StatisticalAnalyzer(df)
analysis = analyzer.analyze_all()
comparison = analyzer.hypothesis_tests(
    "group", "outcome", test_type="auto", bootstrap_samples=0
)
insights = InsightEngine(analysis).get_summary()

print(analysis["descriptive"]["outcome"])
print(comparison["test"], comparison["p_value"])
print(insights["total_insights"])

report = ReportGenerator(analysis, insights, hypothesis_results=comparison)
report.to_json("analysis.json")
report.to_html("analysis.html")

analyze_all() returns sections named overview, descriptive, normality, outliers, correlation, missing_data, data_quality, distributions, column_roles, column_types, histograms, and analysis_warnings. Group comparisons are requested separately; pass their results to ReportGenerator to include them in reports.

More examples

From a repository checkout, run the complete feature showcase to see printed output for every public workflow and generated report files:

python examples/example_usage.py --output-dir reports

Add --skip-interactive if you want only JSON, CSV, and static HTML. For your own data, replace the sample DataFrame with pd.read_csv("data.csv") and select the appropriate group and outcome columns.

Data and statistical limits

• Input must be a nonempty DataFrame with unique, nonempty string column names. Missing values are allowed; unsupported nested, complex, or non-finite numeric values raise InvalidDataError.
• The analyzer copies its input. Undefined or skipped analyses appear as None or in analysis_warnings. Review these warnings before interpreting output.
• Group tests are for independent observations. The automatic choice uses normality and variance screens; it cannot establish that a study design or statistical model is appropriate.
• Chi-square association requires expected counts of at least five in every cell. Bootstrap intervals are exploratory and do not adjust for multiple comparisons.
• Column type and role suggestions are advisory; they do not alter the input or choose analysis columns.
• Processing is in memory. The interactive HTML loads Plotly JavaScript from a CDN when opened in a browser.

Documentation

• API reference: public methods, parameters, return values, and errors.
• Examples guide: complete runnable showcase and output files.
• Changelog: shipped changes.
• Roadmap: product goals and current status.

Development

Install development dependencies and run the checks used in CI:

python -m pip install -e ".[dev]"
python -m ruff check src tests
python -m ruff format --check src tests
python -m mypy src/pyautostat
python -m pytest -q --cov=pyautostat --cov-report=term-missing --cov-fail-under=90

Issues and contributions are welcome through the GitHub issue tracker. Include a small reproducible DataFrame when reporting a data-handling problem.

License

PyAutoStat is distributed under the MIT License.