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

Re: Wukongopterus and Darwinopterus

On Nov 27, 2009, at 5:26 PM, David Peters wrote:

Every right tree we hypothesize will admit new taxa as they are discovered with no change to the rest of the tree topology.

That actually does not follow. It is completely possible to have a "right tree" (one in which all the nestings are accurate if one knows the entire tree of life through omniscience) and then end up with a "wrong tree" as you add new taxa, only to get a "right tree" again as you add still more. If our search algorithms were perfect (which they're not), and our model of evolution were perfect (which it isn't) then we'd still only be guaranteed the right topology when all taxa were known. If anything is missing (which they always will be) there is always the chance that the tree is "wrong" despite having done everything else correctly. Fortunately, we can still get trees that are good enough to be useful. Phylogenies do not need to be 100% accurate to have utility - if they had to be perfect, then phylogenetics would be hopeless.

You're never going to figure out the perimeter of England because you come up against fractals.

Actually, there are ways of integrating to get fractal perimeters (it's just that the answer isn't a single number), but that's a digression.

This complexity you're suggesting is another facet of fractals. Avoid it or delve into madness. All we're looking for are simple models for complex processes taking millions of years and just as many generations.

Just simple enough; not too simple - correlated characters have an impact on topology. Therefore, they need to be considered. If you choose to ignore them, that is still a methodological choice. In some cases, this will confound your results.

The list of apomorphies is enormous!

Doesn't mean it's wrong.

Presuming you mean everything should be unordered and ignored for correlations, there are quite a few manuscripts in the cladistics literature that suggest otherwise.

Posting your question back to you: how did they know they were more right than wrong when they concluded?

Easy: simulation studies know the answer before they start. Therefore, you have the on true tree in hand with which to compare.

Mike, I hate to break it to you, but you like kinda like your parents and your grandparents. If you have a tail or fins or horns or gills, you hide it well.

That's just inheritance, not parsimony. Parsimony would say that gill loss in tetrapods should be a simple stepwise process, but it isn't.

Rare? News to me again. And yes, obviously the shared ones (i.e. synapomorphies) are particularly interesting, but I'm not sure where you're going with this.

Please name any apomorphy that is not rare. By definition any apomorphy will be rare.

Wing reduction in flightless birds. Has occurred convergently dozens of times, mostly in distantly related groups (and often represented by a single species). Paedomorphosis in salamanders also comes to mind. And mouth-brooding in frogs and fish.

Is this based on a particular simulation study? Obviously, if a tree is very large, and contains six or seven typos, then the errors are a smaller percentage of the dataset than in a case where the same six or seven typos exist in a small dataset. However, I'm not sure what this means for "sustaining" typos, or how a matrix with many errors would still be considered healthy.

Typos are to be avoided, but they don't ruin trees of sufficient size, as you indicate and agree with.

Actually, all I said is that the typos will be a smaller percentage of the data in a big matrix - that doesn't mean they have a limited effect on the tree. Not every cell in a matrix has equal impact on the resultant topology.

Actually, they're signs of uncertainty, which is a pretty basic aspect of science, and therefore a common issue. Yes, sometimes such terminology indicates a cop out. Often, it is just being realistic. (Are bats closely related to primates? Well, the *might* be, as some trees support that - but most trees these days don't, so it's *likely* they do not - this is just a realistic assessment of the situation).

Good example. Test it. I have. I know the answer. AND YET, if someone else tests it and comes up with a better answer, I will gladly fold and declare a winner. That's science.

You know *an* answer. As you already suggested, a better tree could come along. Therefore, you *might* have the best tree. In fact, there may be trees already out there that are better than yours as it is. What you have is the answer to the question "what is the most parsimonious nesting pair diagram that can be produced from this given matrix?" We then presume that the answer to the aforementioned question is a decent proxy for the answer to the question "how are these organisms related". That is, however, and assumption, which caries a range of caveats. Incidentally, I helped build one of the largest bat trees (actually, one of the largest mammalian trees) ever constructed. I do not presume to know *the* answer.

--Mike H.

Michael Habib
Assistant Professor of Biology
Chatham University
Woodland Road, Pittsburgh PA  15232
Buhl Hall, Room 226A
(443) 280-0181