Yesterday in my conversation with David Miller I told him I didn't think that the new autosomal "ancestry" tests really delivered the extent of separability of ancestral quanta that most people really expect. Well, look at this principal component chart that Dienekes put up from a paper published by Dr. Mark Shriver. Granted, these tests are good at differentiating continental races...but many people want to know if they are "pure Hakka" or "pure German," as opposed to Cantonese whose ancestors picked up the Hakka dialect or Slavic Sorbs who were Germanized. In terms of the autosomal tests I don't see that they are are at such a level of precision. As I've said, most of the autosomal tests will tell most people just what they see in the mirror every morning (all bets are off you are Brazilian of course).