Validation of clustering results: Percentage of surnames in cluster by reference CEL group.

The table shows clustering results on the global network filtered at k = 150 and NP = 0. The columns represent the largest 20 clusters and the rows the CEL groups in the diagnostic list, while the rows are a selection of 23 CELs with higher values in the table. The cell values are the percentages of unique surnames within each cluster that matches a particular CEL group in the reference list ( = >50 highlighted in bold). Percentages are rounded to the nearest integer and zero values are not shown. The largest 20 clusters shown here account for 5,554 surnames out of a total of 5,787 surnames assigned to 82 clusters. The last row lists the most probable CEL allocation (or CEL combination) to each cluster based on the highest percentages. For example, cluster 4 is 86% Chinese, while cluster 9 is 68% Greek, while cluster 6 is 100% Dutch (see table 2 for a description of CEL codes).