Candy et al. : Dividing population genetic distance data by partitioning optimization 
53 
A 
B 
C 
Number of groups (k) 
Figure 6 
Clustering of genetic distance data for populations of Chinook salmon ( Oncorhynchus tsliawyts- 
clia) from the west coast of Vancouver Island. (A) Hierarchical group membership corresponding 
to minimized cost function (CF) by using bi-partitioning optimization using restrictive growth 
strings (bi-PORGS). Populations that show genetic affiliation but are outside the geographic 
region are denoted by < >. (B) Observed and expected values from the reference distribution 
plotted against k groups for the Chinook salmon data. (C) Gap statistic showing that the first 
optimum grouping occurs at k = 4 and the second optimum grouping occurs at k = 9. Black, 
gray, and white represent status of the populations for a given cluster number. Black and 
white represent populations involved in partitions for a particular value of k , whereas gray 
populations were not involved. 
Weir and Cockerham’s (1984) co-ancestory coefficient 
6 . A number of alternative distance measures could be 
tested with this method, but an examination of these 
measures is beyond the scope of this article. Also, deter- 
mination and comparison of the optimal number of 
groupings directly from the multilocus genotypic data 
(e.g., Pritchard et ah, 2000), instead of from the distance 
measures used here, would provide useful. 
Unlike other clustering methods, PORGS does not 
have to embed distance data in vector space (i.e., mul- 
