SPR and RF distances involving trees with polytomies

[Daniel Casali]

Hi everyone,

TNT calculates SPR and RF distances among trees even if they contain polytomies, but this is not allowed in R package Phangorn (function treedist). It seems that RF distances are not apropriated when the trees have polytomies. About SPR distances - It is fine to use it to calculate distances when one or more trees have polytomies? And the RF results in TNT, in this case, have any meaning?

Best

[EAscarrunz]

The original definition of RF was for fully dichotomic trees, but it's easy to generalise for trees with polytomies. I think Penny & Hendy were the first to do this. The resulting generalised measure is still meaningful, but it takes into account time topological disagreement *and* differences in the degree resolution of the trees. You may want to quantify topological disagreement and resolution separately; quartet-based measures are good for that.

It may be better to call this quantity a "bipartition difference" instead of RF, to be strict with the definitions.