Re: New paper on Neoaves

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

> You and David hold to the assumption that homoplasy is random, and can be
> overcome simply by enlarging the dataset or expanding the taxon sample.

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.

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

Shannon M. Hedtke, Ted M. Townsend & David M. Hillis: Resolution of 
Phylogenetic Conflict in Large Data Sets by Increased Taxon Sampling, 
Systematic Biology 55(3), 522 -- 529 (June 2006)

No abstract, so I'll retype the conclusions:

"No particular number of genes or taxa will guarantee that phylogenetic 
reconstruction is accurate, even if bootstrap support for that 
reconstruction is high. If conflicting signals between genes are due to 
method inconsistency, adding more genes may lead to increasing support for 
the incorrect phylogenetic reconstruction. [That's the definition of "method 
inconsistency".] In such cases, increasing taxon representation may improve 
accuracy more than does increasing gene number. If we incorporate our 
understanding of sources of inconsistency into study design, resulting 
phylogenies are more likely to be representative of evolutionary history.
   For any given study, how can an investigator know whether it is better 
to add more characters or add more taxa to a phylogenetic analysis? High 
support values for individual clades indicate that sufficient characters 
have been collected to converge on a robust result. Unfortunately, the 
well-supported result may be wrong, particularly if small trees with long 
branches are being estimated. This outcocme appears to be especially likely 
when intensively sampled genomes have been selected across relatively few, 
distantly related species -- as with model organisms. In such cases, any 
slight systematic bias can become magnified and misinterpreted as 
phylogenetic signal. High bootstrap or other support values are almost 
guaranteed with genome-sized character sets: the analyses will tend to 
converge on some answer, even if the answer has more to do with biases in 
the analysis than phylogenetic history. Therefore, it is important to 
investigate possible sources of systematic bias, such as long-branch 
attraction or model misspecification. Simulation studies can help determine 
the likelihood of long-branch attraction problems in these situations and 
suggest where additional taxon sampling should occur."

> Here's one case that highlights the problem.  Naylor and Brown (1998) used
> over 12,000 bases (from 19 mitochondrial genes) and recovered 100% 
> bootstrap
> support for a clade that comprised vertebrates and echinoderms to the
> exclusion of amphioxus (lancelet).  This topology was recovered regardless
> of the method of analysis.

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.

>  But the thing that is obvious about *morphology*-based phylogenetic 
> analyses is that they are almost always followed by a discussion of which 
> morphological characters (synapomorphies) unite which taxa.  In other 
> words, it's plain to see the identity of the characters that diagnose 
> certain clades.  This rarely happens with *molecular* clades.  Here, the 
> characters are at the level of genes and amino acids, and the structural 
> and functional properties of the sequences are skimmed over.

The reason is obvious: molecular characters all look the same and are thus 
very boring, with the few exceptions of those that can be treated like 
morphological ones (e. g. the unique 9-bp deletion in BRCA1 that is an 
autapomorphy of Afrotheria).

BTW, the "Ontogeny Discombobulates Phylogeny" paper (Wiens et al., Syst. 
Biol, February 2005) has a couple of clades that have high Bayesian 
posterior probabilities and are nevertheless clearly spurious. However, 
these based on are morphological data of paedomorphic salamanders (miscoded 
as if they were metamorphosed adults).