One of the shortcomings of f3 statistics is that they only allow the analysis of two mixture components at a time. I wanted something that could analyze craniometric data the same way that ADMIXTURE analyzes SNP genotype data, with an arbitrary number of mixture components, so I wrote some code that does just that. Right now supervised analysis, with reference specimens as the mixture components, seems to be working well, but unsupervised analysis, with inferred mixture components, is a more difficult problem that I haven’t completely figured out yet.
The first analysis that I ran was with K = 3, and with Steinheim, the La Ferrassie I original, and the Rota specimen as the components. The admixture plot is at the bottom of the post.
Abri Pataud has a lot of Steinheim-like admixture, but also some Ferrassie-like admixture.
Many of the Ainu and Arikara have big chunks of Ferrassie-like admixture.
The Andamanese are completely free of Ferrassie-like admixture, and seem to have the least amount of archaic admixture of any population.
The Anyang specimens have only small amounts of Ferrassie-like admixture.
The Atayal have low levels of archaic admixture.
Australians have large amounts of archaic admixture, most of which is more Steinheim-like, but some of which is Ferrassie-like.
Broken Hill I surprisingly seems to be much more Ferrassie-like than Steinheim-like.
Because the f3 statistics showed the Buriat archaic admixture to be most like Broken Hill I, I concluded that it was non-Neanderthal heidelbergensis. But this analysis shows most of their archaic admixture as being Ferrassie-like.
Bushmen have hardly any Ferrassie-like admixture, but some of them have significant amounts of Steinheim-like admixture.
Cro Magnon I has a lot of archaic admixture, most of it Ferrassie-like.
The Dogon have relatively little archaic admixture.
The Egyptians have very little archaic admixture.
A lot of Ferrassie-like admixture shows up in the Eskimo.
Chinese and Koreans have little to no Ferrassie-like admixture.
A lot of the Moriori have big chunks of Ferrassie-like admixture.
Peruvians have low levels of archaic admixture, and very little of what they do have is Ferrassie-like.
Predmost III and IV have a lot of archaic admixture.
Skhul V is a lot more Steinheim-like than Ferrassie-like.
Tasmanians have a large amount of archaic admixture, the overwhelming majority of which is more Steinheim-like than Ferrassie-like. The same is generally true of the Tolai.
Upper Cave 101 has a very large amount of archaic admixture, a large majority of which is Ferrassie-like.
Vedda have very little archaic admixture, and it’s entirely Steinheim-like.