Estimating ethnicity
And of course, when people ask me if DNA tests are accurate, I know they’re not asking me if the millions of letters of raw genetic code they downloaded are accurate from SNP to SNP. They just want to know if they’re actually 1% Nigerian and 99% Northern European like 23andMe said. In this specific case, yes, I’m sure this is perfectly true and informative. But if you are Japanese and the service reports that you are 0.1% Ashkenazi Jewish, I wouldn’t believe that (these are both real examples from friends).
How do I make these judgments? Well, in the first case, I happen to have read a 23andMe white paper that noted that their simulations indicate that their method returns a 50% false positive rate below 0.1% Sub-Saharan African ancestry against a European genetic background. This means that by the time you are up to 1%, you’re ten times over that still-noisy threshold, so they’re almost certainly detecting real admixture. On the other hand, when it comes to the Japanese person with a 0.1% Ashkenazi background, I know that companies are banking on the finding that many customers really like 0.1% granularity (this is from market research surveys), but scientifically, this level of faux exactitude often results in false positives.
https://razib.substack.com/p/so-you-want-to-test-your-dna?r=u0rd&utm_campaign=post&utm_medium=web
