Methodology
What the atlas measures—and what it does not.
This atlas compares published city-level averages. It does not analyse individual DNA, reconstruct ethnic identity, or infer population borders.
The source is checked daily. This date changes only when published profile values or source-reported n values change.
1 · Data path
From the public workbook to a city profile
The upstream file is AzerbaijanDNA’s public Analytics workbook. The atlas reads only its AUTOSOMAL worksheet—not the Y-DNA worksheet—and groups records by the CITY field.
Locations appear only after at least three rows have a numeric EEA value. For each included city, every available numeric value in each of the five published fields is averaged and rounded to one decimal place.
EEAGEDROSIACAUCASUSEUROPEANMIDDLEEAST2 · Component resolution
Five published fields, not raw K12b output
Dodecad K12b was originally released as a twelve-component calculator. AzerbaijanDNA’s AUTOSOMAL worksheet stores five broader fields, which the atlas reads directly; it does not recompute them from raw K12b output.
The workbook does not explicitly document its twelve-to-five conversion. The grouping below is an arithmetic reconstruction from published result cards: in one example, Siberian 3.03 + Southeast_Asian 1.55 + East_Asian 1.21 equals the reported EEA total of 5.79. Other published examples are consistent with the same grouping.
Siberian + East_Asian + Southeast_AsianGedrosiaCaucasusAtlantic_Med + North_EuropeanSouthwest_Asian + South_Asian + Northwest_African + East_African + Sub_Saharan3 · Similarity calculation
Unweighted Euclidean distance
For cities A and B, the atlas subtracts the same five percentages, squares those differences, adds them, and takes the square root. All five fields have equal weight. Lower values mean the two published profiles are more similar on this five-number representation.
d(A, B) = √[Σᵢ₌₁⁵ (Aᵢ − Bᵢ)²]Worked example from the current snapshot
The Euclidean distance between Bakı and Xalxal is 0.94.
4 · Filters and ranking
One distance rule throughout the atlas
- The reference city is excluded from its own nearest-match result.
- The nearest profile and ranked list respect the active minimum-n filter.
- Sample-strength shading changes opacity only; it never changes distance or rank.
- The distance threshold changes which profiles are shown in range, not how distance is calculated.
5 · Interpretation limits
A profile comparison, not a biological boundary
- City averages describe the submitted, city-linked records in the source—not every resident or a representative population sample.
- Small n values are especially sensitive to individual records and future source updates.
- The metric is not geographic distance, FST, G25 distance, a migration model, or a test of statistical significance.
- Components from admixture calculators are model-dependent statistical constructs; their labels should not be read as discrete peoples or pure ancestral groups.
- Similarity between two averages does not establish ethnicity, kinship, historical direction of gene flow, or political borders.
Sources and reproducibility
Traceable inputs, committed snapshot
The site uses a committed snapshot rather than fetching the workbook while a visitor is viewing the map. This makes each deployment reproducible and lets source changes be validated before they reach the public atlas.