Provenance Tracking

StatQA provides comprehensive provenance metadata for all generated content, ensuring full reproducibility and traceability.

Metadata Structure

Every Q/A pair includes detailed provenance information:

{
  "provenance": {
    "generated_at": "2025-11-19T19:21:28+00:00",
    "tool": "statqa",
    "tool_version": "0.2.0",
    "generation_method": "template",  # or "llm_paraphrase"
    "analysis_type": "univariate",    # univariate/bivariate/temporal/causal
    "analyzer": "UnivariateAnalyzer",
    "variables": ["age"],
    "statistical_tests": ["shapiro_wilk", "jarque_bera"],
    "python_commands": [
      "data['age'].mean()  # Result: 42.5",
      "data['age'].std()   # Result: 12.3"
    ],
    "llm_model": "gpt-4-turbo",       # if LLM was used
    "template_id": "distribution_summary"  # if template-based
  }
}

Benefits of Provenance Tracking

1. Reproducibility: Exact commands and parameters used

2. Quality Control: Track generation methods and models

3. Audit Trails: Full history of analysis decisions

4. Version Management: Tool and model versions recorded

5. Research Integrity: Transparent methodology documentation