MINI LECTURE 16: On the Absurdity of Standard Genetics Reports.

[@marekglowacki2607]

Maybe a remote association but there is a great "Sicialian Scene" in "True Romance" by Tarantino

[@agiambrone]

As someone with provable Sicilian ancestry and who has done both ancestry and 23andme, I can say this: Neither said "X%" Sicilian. Most of the ancestry on both sites said "Italy-South." Ancestry also circled the center of Sicily on their map (not eastern side) which is spot on accurate.

[@Alex_Plante]

In the Old World, ethnicity is a continuum, so the % ethnicity statements should not be taken literally. Principal component analysis, especially when combined with DNA from ancient samples, provides a fascinating insight into ancient population movements. Professors Reich and Krause have fascinating KZbin videos on the subject.

[@behrad9712]

Thank you so much please make video about the this technique great respect from Iran, wish you health and happiness 🙏👌

[@Sebastian-xf8je]

I always see PCA used for genetics data, but wonder if linearity assumptions holds? How do we know the relationship between genes is linear?

[@galenseilis5971]

PCA involves a linear change of basis from the original basis of the data to an eigenbasis. The eigenbasis is selected using the eigenvectors and eigenvalues obtained from the sample covariance matrix. Nothing about this assumes that there is linearity in expectation in the population.

[@marianialvaro7603]

What is the video on how to play with covariance but keep it positive-semi-defined?

[@nntalebproba]

I am making another one.

[@bb-gi9ky]

any practical result to be used for proper presentation of genetic tests?

[@hansfrankfurter2903]

Prof Taleb, what sources do you recommend for beginners in stat/probability? Thank you 🙏🏼

[@onamixt]

Nothing is understandable in the video. We extracted some orthogonal components, aaaaand?..

[@DavidFregoli]

The no true Sicilian fallacy

[@reggaefan2700]

I took one of these tests recently. After watching this video I feel so bad. The test said I was 11% European. I don't look European at all. In fact, I'm the furthest thing from it.

[@samijee]

what do you look like?@@reggaefan2700

[@frankgrimeson]

I wonder what two people have the greatest genetic distance between them. Is it Han Chinese and Sub-Saharan African as in your rescaled graph? Or is there not enough representation from various groups to know yet? I had always heard that there is more genetic diversity within Africa itself than between Africans and anyone outside of Africa. But maybe that's wrong.

[@bogdanlevi]

I think that's because the sample doesn't cover a lot of sub-saharan African populations. There are 4 African populations in the list as far as I can see, and 3 of them are pretty close geographically. So it might be that sub-saharan Africans are underrepresented in the sample, and thus they get clustered together in the corner.

[@frankgrimeson]

@@bogdanlevi If you take a look at the rescaled plot (starting at around the 2:14 mark) you'll see a huge distance between Han Chinese and the existing Sub-Saharan African groups -- spanning pretty much the largest distance possible on the graph (top right corner to bottom left corner). For a new, as of yet untested, African group to have an even larger distance, you'd have to zoom the graph out quite a bit and I'm not really sure where those new points would go.

[@bogdanlevi]

@@frankgrimeson if we include more African groups in the sample, the principal components will be different, and so the graph won't just be rescaled, it will probably change entirely.

[@frankgrimeson]

@@bogdanlevi True the components would change, but not "entirely," otherwise the current plot would be useless. It would almost certainly result in a compression since the closeness of existing groups in the data wouldn't change. If you look closely however, there actually is already a diverse selection of African genomes: we have both West African e.g. Gambia, Mbuti, and East African e.g. Ethiopia, Somalia. I think much of the potential distance is already represented in this graph. You can see that some of those African groups are what really push Europeans to the very right of the plot. Which then still confirms my observation that potentially the Han - African distance is the greatest distance between two extant human groups. Unless there's an as-of-yet-untested African group that would somehow compress the entire current plot to less than half its size.

[@EverydayLinguistics]

I always thought that made no sense! Every ethnicity is ultimately a mix of something, so how can you fall into one category and another in different percentages? What a weird way of handling information. They do what sells, I guess.

[@KazuyaMithra]

The only more accurate assessment is through ancient populations. Someone can be 100% English but can have more ancient French and Wesh people in their lineage than Anglo Saxons.

[@Ruhgtfo]

wow

[@ncroc]

I have seen the dna results of Albanians & Greeks on youtube. Even Albanians from Kosovo are 80-90% Greek :) Apparently, Albanians have got more Greek dna than most Greeks :) or may be just may be It's the Greeks that have Albanian dna :)