From Dienekes:
"Any distance measure that does not consider the evolutionary relationship between components will not estimate the distance of individuals correctly.
Take the following three individuals
Bob: 100% European, 0% African, 0% Asian
Alice: 50% European, 50% African, 0% Asian
Eve: 50% European, 0% African, 50% Asian
If we take the Euclidean distance, we get:
dist(Bob, Alice) = sqrt(0.5^2+0.5^2+0) = 0.71
dist(Bob, Eve) = sqrt(0.5^2+0+0.5^2) = 0.71
We might conclude from this that Bob is equidistant to Alice and Eve. But this is not the case, as African, and Asian are not equidistant with European, and Eve is genetically closer to Bob than Alice is."
Summary: for Admixture K studies:
Root mean square deviation is not a correct representation of the true distance between populations.
Dendrograms not based on evolutionary distances are also incorrect