They say that you should choose the number of clusters that can explain about 80% of the variance when you do K means clustering. However in human genetics, all variance should be explained by the genes but that would generate an absurd amount of clusters.
How many clusters are ideal? Does anyone have any data on how many clusters explain 80% of the variance in the human race?

The criticism that rosenberg got was that he arbitrarily chose k=6 clusters. And that's not good enough. We need to prove that 6 are the best or so.

I posted this thread on /his/ but it's filled with haplogroup autists and anti-abo posters with an agenda