Re: New paper on Neoaves

[Tim Williams <twilliams_alpha@hotmail.com>]

David Marjanovic wrote:

> That was a late-night oversight on my part. If some of the taxa in a data 
> matrix have a base-
> composition bias, an evolution model estimated from the data can clearly 
> lead to a wrong result.

Yes.  But even worse, adding data can actually make the situation worse by 
making the erroneous topology well-supported, and deluding you into thinking 
that the 'fake' clade is 'real'.  The tree looks 'better' (nicer bootstrap 
or posterior probabities) but the topology is 'worse', because homoplasy is 
no longer the 'noise' but the signal.

> Adding taxa seems to be a good idea in general, however.

I general, yes.  Nevertheless, I think we need a method to determine whether 
more taxa are helping the analysis, rather than just assuming that more taxa 
will help.

There's also the matter that improved taxon sampling.  Some lineages make 
life difficult by leaving us only a few lingering species.  Try resolving 
the relationships of hyracoids, elephants, and sea cows, for example.  
Between them they muster about a dozen extant species, and even less genera. 
 (OK, you might get lucky and get your hands on some Steller's sea-cow or 
mammoth DNA; but even then the taxon sampling dries up at 13 species.)  Then 
there's the tubulidentates, perissodactyls, etc., which are also stingy in 
this regard.  I'm not blaming them, of course.  They tried their hardest to 
survive into the Present.  And it's not as though humans made it easy for 
them.

> Since this was 1998 (you know, when some people honestly believed Rodentia 
> was paraphyletic to the rest of Placentalia, and Passeriformes the 
> sister-group to all other extant birds), I'll simply blame the model, and 
> perhaps the fact that Bayesian analysis was not yet available.

No, the "wrong model" isn't the problem here; they tried many different 
models (four substitution, four among-site rate-heterogeneity), which are 
used today.  Bayesian analysis didn't solve the problem either (I know one 
of the authors personally).

> The reason is obvious: molecular characters all look the same and are thus 
> very boring,

To a morphologist, yet.  But people who study molecules beg to differ.  We 
need to get these people on board so we can delve deeper into the sequence 
data and get a handle on what exactly is driving the signal, if not 
phylogeny.

Cheers

Tim