[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

AW: Density of intraclade sampling (was: More evidence for Eufalconimorphae and Psittacopasserae)

> > You need at least 2 of _Sagittarius_, _Pandion_ and
> elanids plus 2 of eagle/hawk/buteo. One of
> _Sagittarius_/_Pandion_ will usually go rogue.
> > 
> >> Isn't that rather strange if one is interested in
> the
> >> position of falcons?

Not at all. _Pandion_ likes to LBA, _Sagittarius_ too but less so. Meaning the 
root has a high chance of being wrong if you use less taxa. If you use for 
example only _Pandion_ and _Accipiter_, you are likely to end up with a wrong 
ancestral state (even if the two *do* clade).

Which will lead to LBA so inconspicuous you don't likely realize it (there are 
no *suspiciously* long branches needed to shift members of an explosive 
radiation). And with parrots often winding up too close for comfort to 
Accipitri, a wrong ancestral state is a significant cause of error.

(Essentially we're not discussing falcons, but the basal radiation of the 
"higher landbirds". Of which falcons, songbirds, parrots, but also hawks, 
woodpeckers, rollers, probably owls, and possibly cypselomorphs and seriemas* 
are all part.)

The key is realizing that these problems will NOT lead to reduced branch 
support**. So to tell whether you have a problem you need to reproduce 
intraclade phylogeny, and this requires 3 terminals at least (1 basal, 2 

(This holds for morph data too. L&Z could easily tell that their "raptor clade" 
was bogus, because the root of Accipitri was in a position relative to falcons 
and owls that was contrary to *all* other evidence.)

As another thing, cladistics is again confirmed as superior to phenetics. You 
*don't* want to see the trees a phenetic analysis gets from these data... the 
cladistic trees are dangerously unstable in certain points; the phenetic 
trees... well they have branches and leaves and a root, but that's about all 
that can be said in their favor.



* I am still very unsure about seriemas. Many "Metaves" have a tendency to be 
unstable between Aequornithes and "higher landbirds". But their effect on 
topology is not very profound (unlike mousebirds). Seriemas might qualify for 
this group of taxa too - they are not major "rogues", but they are major "rogue 
prey" (above-average susceptible to effects of "rogue" taxa).

** I am unsure of the implications of this. An analysis that can be provoked to 
yield 2 mutually exclusive results, both with supports of about 0.9 - shouldn't 
happen, particularly not if the analysis is technically sound. If it were an 
either/or thing - removing/including one taxon favors one sister-group 
relationship over the other... but it isn't. Support of bogus "clades" - either 
the +mousebirds or the -mousebirds version (or both) *has* to be bogus - is 
simply too high.