F3 test for admixture

Question

[abascalfederico]

Hi,

I am trying to understand how the f3 and f4 tests of admixture work.

I've come to a question and hope someone here can help: the f3 test determines whether the relationships between X, Y and W can be explained with a tree (no admixture) or instead need a graph (admixture) to explain the variation of allele frequencies along history (drift).

The f3 test can be written as f3(X; Y,W). If I understood correctly, it works with unrooted trees. If that is the case I think it should not matter the order of the nodes, I mean f3(X; Y,W) or f3(Y; X,W) or f(W; X,Y) should be all equivalent.

However, since the order of nodes is usually specified, I guess I am probably misunderstanding something. Any help is very welcome.

Thanks!

Federico

[stolarek.ir]

By logic if you get a significant f3 statistic, you say, there was an admixture for lets say X population when analysing Y and W, but when you now try to analyze Y admixture with respect to X and W population you kind of change the direction of gene flow.

But still when in doubt I strongly advise, just write an email to the authors of publications, which use those statistics

[abascalfederico]

Thank you stolarek.ir,

What you point out about the direction of gene flow is intuitive. However, if trees are unrooted there is no directionality (until you root the tree). That's why I am confused. I may finally email the authors. Thanks again