Motorcycle Mama

I would very much like to be informed of the progression of the state of the art, without a great deal of detail please.

Tom Ames

One obvious advance is the introduction of the bootstrap value to evaluate confidence in a given hypothesis.

Although the non-parametric bootstrap can be applied equally well to morphological and molecular data, the parametric boostrap is uniquely suited to data for which there is a model of evolution. For now, this mainly means molecular data.

Phylogenies were once upon a time grossly misused. But once systematists realized that they needed to be testing specific hypotheses, this situation improved.

The cladists (what's left of them) still disagree with this approach (that is, of treating systematics as a framework for statistical hypothesis testing.)

Motorcycle Mama

Thanks,
I was aware of the bootstrap as well as the production of consensus trees. Your comment about testing hypotheses is interesting since one of the mantras of the cladists in the 80s was just that, a cladogram is a hypothesis, etc., etc. But then that's always been implicit in systematics, a taxon is a hypothesis.
But to return to my original point of vagueness in molecular data, I've seen the results of bootstrapped and consensused molecular data that was still ambiguous, i.e., there were more unresolved than resolved branches. I would never claim that morphological, or any other kind of data, are immune to such vagueness but what I take exception to, as I believe I've said in previous responses, is that the molecular data is afforded some special status.
MM

Tom Ames

This is not quite true. One of the cladist mantras is that parismony "minimizes the number of ad hoc hypotheses of homoplasy".

They would tell you that a tree with 514 "ad hoc hypotheses of homoplasy" was preferable to one with 515. Implicit in this reasoning is that there are no confidence intervals on the estimates of branch length. (Furthermore, it's known that parsimony is positively misleading under certain quite common conditions -- the famous "long branch-length attraction" of the "Felsenstein zone".) A tree, to the cladists, is specifically not a statistical hypothesis: they'd say that it's an assumption-free portrayal of the weight of the evidence.

Well, I've seen lots of statistics that do not include confidence intervals, or that lead to an ambiguous conclusion. So what? We test those hypotheses that the data can address, and leave the others alone. This is not a function of the type of data. It's a function of the use to which the data is put.

You've said that molecular data are afforded a special status, and I outlined the ways in which you are correct. However, you still seem to think there is a quasi-mystical aspect to this. What is comes down to is:

<ol type="1">[*]molecular data are better suited for some kinds of analyses[*]molecular data are more more objective in some ways[*]molecular data are more easily collected.[/list=a]

Why is this a problem?

Oolon Colluphid

Sorry, coming in a bit late here...

Surely the most obvious thing that DNA data has over morphological analyses is that we know that the patterns of DNA are passed down through generations with remarkable but not perfect fidelity? Even into separated lineages. By taking evolution to be true, we expect the most closely related organisms to have the most similar DNA; conversely, the fact that organisms thought closely related for many many other reasons do indeed have similar DNA confirms evolution, again.

Cheers, Oolon

Motorcycle Mama

Look, forget it. We seem to be arguing at cross purposes. My point is that the ambiguity in molecular, just as in morphological, data has to have an explanation, an explanation that is not yet forthcoming.
MM