LIMITATIONS.md

Limitations, ethics and responsible interpretation

1. No validated education score: Eight indicators are not a complete education system assessment or an official SDG4 scorecard.
2. No causal inference: Associations between spending, technology, access and outcomes are descriptive and potentially confounded. No panel regression, causal design or statistical uncertainty interval is supplied.
3. Not all observations are comparable: Gross enrollment and primary completion ratios may exceed 100. Years, definitions, collection quality and measurement methods vary.
4. Inexact latest-year ranking: The overview ranks latest reported values that can have different reporting years. It is not necessarily a common-year league table.
5. Missing information stays missing: Low coverage is visible; a missing point is not zero and is not estimated. In some countries some indicators have very sparse coverage.
6. No unseen national/individual inequality inference: The Country gaps view measures differences between two national aggregates on exactly the same year. It cannot reveal gender, income, disability, ethnicity, geographic or individual-level gaps.
7. Descriptive trends only: Changes in reported levels do not establish intervention effects, growth-adjusted improvements, learning achievement or welfare effects.
8. Reported dates and source revisions: Data retrieved today may later be revised by the World Bank or source producers. Download sessions and cite retrieval dates for reproducibility.
9. Data provider rights: Repository MIT licensing covers original code only, not World Bank/UNESCO source datasets.
10. Synthetic demo: Fictional values are generated for interface testing, never to be used as evidence or publication data.
11. API availability: Live API availability and browser cross-origin access are not guaranteed. This build has mocked local API integration tests; a production connection check is needed after deployment.
12. Study design: For a scientific comparative education paper, preregister selection, inspect indicator definitions, assess sample sizes and data vintage, and perform robust contextual and inferential analyses outside this tool.