By using Ward’s clustering algorithm based on the spectral embedding, all but 81 of the European sample were clustered into one of eight relatively large European clusters (labeled A–HH, Table I). Figure 5 illustrates the distribution of country groupings within each cluster (i.e., the conditional probability of country grouping given cluster). Clusters tend to consist of individuals sampled from a common ancestry, or likewise, people of common ancestry tend to be clustered together. Labeling the resulting clusters in Figure 5 by the primary source of their membership highlights the results: (A) Swiss, (B) British Isles, (C) Iberian Peninsula, (D) Italian A, (E) Central, (F) Italian B, (G) North East, and (H) South East. The remaining four small clusters show a diversity of membership and are simply labeled I, J, K, and L. Cluster L has only seven members who could be classified by country of origin and is not considered further.
