Responsible use guidelines · Known limitations · Data quality framework
All datasets are compiled from publicly available official sources (WHO, PAHO, national statistics institutes) under open access policies. No individual-level data is included — all variables are aggregate population-level indicators. Personal privacy cannot be compromised by these datasets.
These datasets are designed for academic research, policy analysis, public health planning, and educational purposes. They are NOT intended for: individual clinical diagnosis, commercial targeting of individuals, discrimination based on demographic characteristics, or any application that could harm the populations described.
1. Temporal gaps: Some countries have missing data for specific years due to conflicts, political instability, or incomplete reporting to international organizations. Missing values are coded as NA. 2. Source heterogeneity: Despite harmonization efforts, some variables may not be perfectly comparable across countries due to differences in national definitions. 3. Ecological fallacy: Aggregate-level associations do not imply individual-level causation. 4. Projection uncertainty: Future projections assume continuation of observed trends — structural shocks (pandemics, conflicts, policy changes) are not modeled.
Data quality is flagged at three levels: High (direct from official primary source, complete coverage), Medium (estimated or interpolated values, partial coverage), Low (proxy indicators or older data used due to absence of direct measures). Quality flags are included in each dataset as the 'data_quality' variable.
Countries with stronger statistical systems (Chile, Uruguay, Argentina, Spain) tend to have higher data quality ratings. This may create apparent differences in indicators that partly reflect measurement quality rather than true population differences. Users should interpret cross-national comparisons with this limitation in mind.
All analysis code is available in /analysis/ folders. All datasets are versioned with DOIs. All transformations are documented in METHODOLOGY.md. We commit to maintaining these resources and responding to reproducibility queries within 14 days.
If you identify an error, inconsistency, or ethical concern: open a GitHub issue at github.com/juanmoisesd or email juanmoises.delaserna@unir.net. All reports are reviewed within 7 days and documented in ERRATA.md.