Quote:
I was looking at everyone's results and something is VERY VERY OFF here because Eurasians (Japanese and Europeans) should be MUCH closer to each other than Eurasians to Yoruba (even with basal eurasian factored in) and not the other way around. Reason is
because ALL Eurasians share substantially the same grandparents from the Out-of-Africa event.
The G25 can't be that bad. Try using SCALED to see if that fixes the problem and Europeans become significantly closer to Japanese than Europeans to Yoruba
Those coordinates don´t work
only giving more % to those references which have closer distance,.
With this drawing you can better understand how these calculators work, which use principles based on multivariate statistics.(it´s only to have an aproximate intuitive idea about how it works, don´t take this lyerally)
https://i.imgur.com/Z1FtCBz.png
A,B,C,D,E...= references in sources
A´,B´,C´,...=distances to every reference.
1,2,3,4= possible solutions, it´s suposed that all of them are projected always passing through the closest reference.
The algoritm would chose a solution passing through(or near inside the dispersion cloud) A,B,D and E references and would calculate the % according an specific configuration.
F is closer to you than D, but is out of your admixture and this component would no score any % in your results, because it´s close to you, but not to the most probable projection line that would be the solution.
All this is quite improvised, but I thonk it´s a graphic and intuitive way to explain something complex as this issue.
We can see that it works in this way, watching results and distances in many model, and for example, Spaniards usually score in NA , although NA references have greater distances than many others, including other non European references.(I explain it with this example because I am Spanish, and NA in Spanish is a recurrent topic, nothing else)
