The Study of Racialism

mr sweet, you cite 30% as the percent of white americans who have recent african dna. I have seen several estimates ranging from 20-30%.

But what do you think about this washington post article that cites a 5% figure?

http://www.washingtonpost.com/wp-dyn/co ... v=hcmodule

According to Mark Shriver, professor of genetics and anthropology at the university, DNA testing reveals that 5 percent of white Americans have some African ancestry and 60 percent of black Americans have white bloodlines.

Let me pick two minor nits before answering. First, this topic belongs under "Molecular Anthropology and Genetics", not here. "Forum Management" is for discussions of the mechanics (software, downtime, how-to) of the forum itself. Accordingly, I am moving this thread. Second, it is not strictly correct to say "According to Mark Shriver" when in fact you are citing an non-peer-reviewed popular press article that claims to quote Mark Shriver. Better to say "according to Ralph W. Eubanks, Mark Shriver said,...."

Okay. Let me try to answer you.

To see why your question has no single best answer, look at the following graph.

The above graph was prepared by then-grad student Heather Collins-Schramm at the U of California, Davis. The four clusters of dots reflect Afro-European admixture measurements taken from four ethnic groups: 147 European Americans, 264 African Americans, 135 Africans from Zaire, and 159 Africans from Nigeria. We are interested in is the leftmost group, the 147 White Americans.

As you can see, most of the 147 European Americans are jammed up against the top of the graph (the top represents 100 European, 0 percent African). This means that most of them are overwhelmingly European. But we are interested in how many White Americans have even the slightest trace of African admixture. Your question comes down to asking, "How many of those 147 White dots are NOT on the top line. How many are even the tiniest bit below the line?"

When Steve Sailer asked Mark this question in May 2002, Mark answered that about 30% of those dots have at least 2 percent African ancestry (and 98 percent European). But admixture ratios of less than 5 percent are meaningless--today's tests are simply not that accurate yet. With this in mind, when I spoke to Mark in November 2003, I counted dots and found that only about about 10 percent of the white dots have less than 95-5 Euro-African mix. Since then, on the one hand admixture mapping has improved in precision and accuracy but on the other hand, we have learned to be more cautious in dot-counting. The figure quoted in the article that you cited comes from counting only those dots that have 7 percent or more African admixture--that is, where we can be 95 percent certain that the test is detecting something real.

And so, how many White Americans have African ancestry depends on how much ancestry must be found for it to "count", given that all tests are imperfect. If you mean any marker at all, then the answer is thirty percent. If you mean enough markers to exceed the error limits of the test, then it is about ten percent. But if you mean with 95% certainty, then the answer is five percent.

I would look at it differently. I would not ask how many White Americans have "significant" African ancestry for two reasons. First, as I just explained, "significant" is in the eye of the beholder. Second, White Americans come in many flavors and the differences among them are profound.

A better questions is, "How much African admixture is in the U.S. White population?" This answer, 0.7 percent, you can take to the bank. An even better question is, "How does African admixture vary among different White groups?" White Puerto Ricans (like my mom, say) can run as high as 50% African. Lumbees (like Heather Locklear) are about 35 percent African. And Louisianans, East Texans, and Melungeons run about 5 percent African.

thanks for your detailed reply.

Your software allowed me to post only in that forum management forum.

As for the test data, you are saying, if I am correct, that the tests show that 30% of white Americans do have recent African heritage, but that 30% figure is outside of the 95% confidence interval for the test?

Is there are a possibility in the near future that this test result of 30% may rise in confidence?

Why do you say 30% if the data does not support it?

That was before you were assigned posting privilege. It sometimes takes a few hours.


That is correct.


I think so. The labs that do this sort of thing keep finding new markers and the more markers, the more reliable the measurement. The number of known continent-of-ancestry markers have quintupled since Mark wrote his first study.


In a nutshell, that was then, this is now. If you are refering to Afro-European Genetic Admixture in the United States, then, as the footnote (19) states, I was quoting Sailer who was quoting Shriver. I would word it more cautiously if I were to write the same essay today. I suspect that Mark also words things more cautiously today. Again, we are looking at the outlier tail of a very skewed distribution, so it is better either to look at the overall number (0.7) or to look at each sub-cluster of dots (Creoles, Lumbees, Melungeons, etc.).

I don't think I'd go so far as to say they're meaningless. In fact, on AncestryByDna's website, it is mentioned that even results below the 95% confidence interval should be taken into account and not just summarily dismissed. The confidence interval may diminish to 80%, for example, but this is still telling us something useful. Also, I have seen studies where results have been been below or equal to the confidence interval, but the researchers have accepted the results as proof of the presence of a certain marker (and, consequently, proof of admixture), without worrying about quantifying. This, I believe, can be done with impunity. Please let me know if you think I'm way off base. Here's an example in a blood protein study from 1991:



http://www.ncbi.nlm.nih.gov/entrez/quer ... ds=2021190




Do you mean "more" rather than "less"?




I think all researchers have different opinions on how much needs to be found to "count." Interestingly, AncestryByDna mentions that certain populations consistently come up with admixture levels below the 95% confidence interval. The fact that the results are consistent and replicable carries quite a lot of weight, in my opinion and that of AncestryByDNA. If one were to dismiss these results as useless merely because of the confidence interval, I think one would be underestimating admixture. Heck, even the 95% confidence interval is by its very name and nature still not a foolproof 100%.

This is just my two cents, which doesn't really mean anything, since I'm just an ordinary Joe Schmoe who happens to have an interest in this stuff!

Yes. I meant "less than 95% European but more than 5% African".


I think you hit the nail on the head. It all depends on what you are looking for. If you are trying to estimate the annual rate of Black-to-White "passing," then you will want to be conservative and go with a smaller figure like 5-10 percent. But if (like cry0fan) you are trying to change society by forcing Americans to recognize the red-flag idiocy of "race" for what it is (a way of maintaining class inequality), then any replicable amount should count, and you should go with the original 30 percent figure.

I should reword what I said above to clarify:

Interestingly, AncestryByDna mentions that certain populations consistently come up with measurable admixture levels that always happen to be below the 95% confidence interval. The fact that the results are consistent and replicable carries quite a lot of weight, in my opinion and that of AncestryByDNA. If one were to dismiss these results as useless merely because of the confidence interval, I think one would be underestimating admixture.

I think you knew what I meant, anyway.