CLUMPP can be used on multiple runs of the Structure software to match up the labels of clusters that correspond to each other, but is it possible to use CLUMPP to match clusters at different K levels?
Obviously there's potential for mistakes, but I can think of scenarios where at K=2 samples are clustered into African vs non-African, and at K=3 the non-African individuals divide into two clusters. This would suggest a hierarchical clustering, wouldn't it?