Re: SV: 55 million year old parrot found/(Number-)Crunching _Mopsitta_ et al

[David Marjanovic <david.marjanovic@gmx.at>]

> On the subject of cladistic analyses: I am still pondering the issue of 
> cladistic analyses of fragmentary material. If anyone can recommend 
> background, theoretical or case studies that would enable one to make a 
> cladistic analysis of taxa with an usually insufficient hypodigm which 
> yields results that are at least *somewhat* resolved.

The entire first half of the June 2003 issue of JVP.

> For example: _Graculavus_ looks very much like it's a charadriiform and 
> for the sake of simplicity I presume that this is not falsified in a 
> preceding analysis. Now, taking a selection of taxa encompassing the 
> morphological and phylogenetic diversity of the Charadrii, of the 
> Scolopaci, of the Lari, etc, and adding _Graculavus_, and only scoring 
> characters than are present in _Graculavus_: would
> a) clades resolve according to the existing consensus and would
> b) _Graculavus_ tend to join a particular one of these, or would it float 
> around aimlessly at the charadriiform base?

Completely impossible to predict. This is a purely empirical question.

However, don't bother. _Don't_ only score characters that are present *G.*. 
Instead, score _every single_ character that is parsimony-informative for 
your taxon sample*. Total-evidence approach. Otherwise you risk getting 
grossly spurious topologies. For example, teeth alone will give you 
"Ungulata" and will find you whales and mesonychians as sister-groups... 
Sure, the focus of your interest is *G.*, but it's not the focus of your 
computer's interest. Your computer needs the whole dataset.

* Ideally after having found out which ones contain phylogenetic signal... 
but that's probably not necessary, because random noise is random. The 
signal adds up, and the noise cancels itself out. Just make sure you don't 
use correlated characters.

> And there can be other benefits. For example the analysis run in the 
> description of _Piksi_ pointed out a technically possible scenario that 
> remains untested: that _Telmatornis_ is a member of the Mirandornithes 
> predating the flamingo-grebe split. And that therefore any study of 
> _Telmatornis_ is very well advised to include grebes and the most ancient 
> Phoenicopteriformes.

In other words, a phylogenetic analysis of the position of *T.* cannot help 
being an analysis of the whole of Neornithes or at least Neoaves? Doesn't 
surprise me...