Below is the unsupervised K = 2 admixture plot for Howells’ craniometric data. I expected the K = 2 analysis to produce an archaic component and a modern component, and that’s exactly what it did.
Some Eskimos, Polynesians, Australians, Melanesians, Tasmanians, and Africans actually have more of the archaic component than the archaic fossils do. The only possible explanation for this is erectus admixture.
La Ferrassie I, La Chapelle, Cro Magnon 1, Upper Cave, Steinheim, Djebel Irhoud 1, Predmost III, and Shanidar 1 all have more of the archaic component than Broken Hill I does, which strongly suggests that Broken Hill I was very close to the ancestor of modern humans.
The Rota specimen has a low level of the archaic component, but it’s not as special as it seemed from the population-level f3 statistics. There are many individuals from several populations that have less of the archaic component than the Rota specimen does.
Remember that, as with ADMIXTURE, the estimated admixture proportions may be different from what they actually are. The results provide an accurate picture of the relative levels of admixture, but not necessarily the absolute levels of admixture. For example, the Eskimo at the top is not really 99% archaic; it just had the most archaic cranial morphology of any individual in Howells’ data set.