Continuing the post on MCLUST analysis of the Howells' dataset, below is the Mahalanobis distance matrix between the centroids of the 14 clusters:
The most striking feature is, of course, the great morphological distance of the Neandertal cluster #6 to all modern human groups, which exceeds the maximum distance between modern human groups (Bushman vs. Moriori/Maori).
My gestalt impression is that the distances here don't really seem to have much to do with geographical proximity, or autosomal FSTs either from what I can remember of them. Particularly it is my impression the groups we refer to as "continental races" - i.e. Negroids, Amerinds, Mongoloids (using East Asian as proxy for Mongoloid), Caucasoids - have relatively close distances with one another with the outliers to this seeming to be Lateral Caucasoid->East Asian, Lateral Caucasoid->Negroid and Amerind->Negroid.
The Australoids and Santa Cruz (and Maori?) relative closeness to Neanderthal cluster seems like it maps an overall robusticity.
Might I ask if the choice of the Linear/Lateral labels reflects or maps to a historical physical anthropology distinction/categorisation?
Yes, in traditional anthropological terms Berg is "Alpine", i.e., broad
Also, the mean of cluster #1 in GOL (head length) is +0.39SD and of #7 it is -0.42 and of XCB (head breadth) they are +0.15 and +1.02 respectively, and of ZYB (face breadth) -0.44 and +0.12.
So, #1 is long-headed, narrow-headed and narrow-faced on average whereas #7 is short-headed, broad-headed and medium-faced. Note that the face breadth of the "Alpine" #7 is only +0.12 because the greatest human breadths are found among Mongoloid groups, e.g., #12 (Buriat) has +1.48SD face breadth.
Also, yes, the major groups are relatively close to each other. The greatest morphological distances are found between people on the fringes (Eskimos, Buriats, Polynesians, Bushmen, etc.)
Maybe it would be better to group lateral and linear Caucasoid into a single Caucasoid group and redo the analysis that way. It doesn't look right, like if it's an artifact, not something real. Sort of like Rosenberg's Kalash cluster in his 2005 study, which disappeared when he used more SNPs. The other clusters seem ok.
Maybe it would be better to group lateral and linear Caucasoid into a single Caucasoid group and redo the analysis that way.
The whole point is that the algorithm discovers the clusters for itself, I don't "group" anything myself.
It doesn't look right, like if it's an artifact, not something real. Sort of like Rosenberg's Kalash cluster in his 2005 study, which disappeared when he used more SNPs. The other clusters seem ok.
I can confirm that the Kalash cluster does appear when using lots of SNPs. It's no artifact, probably just a result of that population's isolation.
Stay on topic. Be polite. Use facts and arguments. Be Brief. Do not post back to back comments in the same thread, unless you absolutely have to. Don't quote excessively. Google before you ask.
My gestalt impression is that the distances here don't really seem to have much to do with geographical proximity, or autosomal FSTs either from what I can remember of them. Particularly it is my impression the groups we refer to as "continental races" - i.e. Negroids, Amerinds, Mongoloids (using East Asian as proxy for Mongoloid), Caucasoids - have relatively close distances with one another with the outliers to this seeming to be Lateral Caucasoid->East Asian, Lateral Caucasoid->Negroid and Amerind->Negroid.
ReplyDeleteThe Australoids and Santa Cruz (and Maori?) relative closeness to Neanderthal cluster seems like it maps an overall robusticity.
Might I ask if the choice of the Linear/Lateral labels reflects or maps to a historical physical anthropology distinction/categorisation?
Yes, in traditional anthropological terms Berg is "Alpine", i.e., broad
ReplyDeleteAlso, the mean of cluster #1 in GOL (head length) is +0.39SD and of #7 it is -0.42 and of XCB (head breadth) they are +0.15 and +1.02 respectively, and of ZYB (face breadth) -0.44 and +0.12.
So, #1 is long-headed, narrow-headed and narrow-faced on average whereas #7 is short-headed, broad-headed and medium-faced. Note that the face breadth of the "Alpine" #7 is only +0.12 because the greatest human breadths are found among Mongoloid groups, e.g., #12 (Buriat) has +1.48SD face breadth.
Also, yes, the major groups are relatively close to each other. The greatest morphological distances are found between people on the fringes (Eskimos, Buriats, Polynesians, Bushmen, etc.)
ReplyDeleteMaybe it would be better to group lateral and linear Caucasoid into a single Caucasoid group and redo the analysis that way. It doesn't look right, like if it's an artifact, not something real. Sort of like Rosenberg's Kalash cluster in his 2005 study, which disappeared when he used more SNPs. The other clusters seem ok.
ReplyDeleteMaybe it would be better to group lateral and linear Caucasoid into a single Caucasoid group and redo the analysis that way.
ReplyDeleteThe whole point is that the algorithm discovers the clusters for itself, I don't "group" anything myself.
It doesn't look right, like if it's an artifact, not something real. Sort of like Rosenberg's Kalash cluster in his 2005 study, which disappeared when he used more SNPs. The other clusters seem ok.
I can confirm that the Kalash cluster does appear when using lots of SNPs. It's no artifact, probably just a result of that population's isolation.
"The greatest morphological distances are found between people on the fringes (Eskimos, Buriats, Polynesians, Bushmen, etc.)"
ReplyDeleteThat's not at all surprising as it has long been recognised to be the case in virtually all species and groups of species (genera).