You certainly have a point. I'm not as qualified as someone who actually holds a degree in biology - though the quote you refer to wasn't about biology. It was about scientific method and the philosophy of science.
Fair enough, though I think a good understanding of a specific field never hurts when evaluating claims in that field
When it comes to the details of biology I will always defer to those with an education in biology. But it's not the details I'm disputing. More specifically, I'm not saying genetics is false or that mutation doesn't happen or that dinosaurs never existed or that one can't make a long list of comparisons and contrasts between the different species that pertain to how life interacts. IOW it's not the data I'm disputing, but some of the conclusions.
Feel free to elaborate, or if you've already done so, refer me back to posts (I didn't read the whole thread before I jumped in).
Feel free to chime in, though, no, I don't need statistics explained to me. Further, though you may already know it, "chaos" has some specific mathematical connotations that are distinct from the random. I did plan to bring that up at some point, but for now it can be lumped in with complex determinism.
I don't think I could whip out the correct definition of chaos off the top of my head, but I know it's deterministic. That was the point
(I've been quite fascinated by chaos ever since it popped up in our
IB maths project on population modelling!)
[ASIDE: Incidentally, I'm not entirely sure whether I should view natural selection as deterministic but complex or as "quasi-deterministic", i.e. predictable only in a statistical sense. I brought this up in a thread quite a while ago, and IIRC I got shot down with the argument that there's nothing random about the maths that predicts the evolution of allele frequencies under selection, but on reflection I'm not sure that is a good argument. A selection coefficient, after all, is just a measure of your
chance of reproducing compared to other genotypes in the population. Then again, are the factors influencing your departure from that chance random or just too complex to model? Bah, I have to say population genetics is not my thing. And this whole rumination doesn't really affect the general predictability of selection as far as I can see.]
But I'm still not sure what you're getting at. Would it be #4 in my list - a combination of determinism and randomness? That still leave the question: is there a belief that something purely random is at work - even if it is mixed with other effects? The purely random, by definition, cannot be predicted.
Well, insofar as quantum phenomena are random, pure randomness is at the bottom of
everything. The real world is pretty predictable for something operated deep down by coin flipping
And statistics does not usually posit the purely random.
But it does deal with phenomena that are random for all practical purposes.
A distribution essentially posits something random working through a deterministic filter as it were.
I think that's actually a decent characterisation of evolution by natural selection! Mutation is random
and highly unpredictable, but one of the sorting mechanisms is... if not deterministic, then at least pretty predictable, I think.
(I don't think evolution is highly predictable on a
large scale, for what it's worth. Partly because what traits
might evolve depends entirely on the pool of available mutations, and partly because ecosystems are simply too complex with too many interactions, feedbacks and plain old accidents. I think there is a degree of predictability that is exemplified by convergent evolution - give different organisms a similar selection pressure, and the solutions are often similar in a quite general way, as in swimming fast requires streamlining or powered flight requires wings of some sort. But not to the degree that, say, vertebrates or flowering plants are an inevitable outcome of evolution.)
Interesting. Thanks for the information, and I will check it out.
[edit] Oh, sweetness! Just read the abstract. I'd willingly put down the $32 to buy a copy, but I'm afraid I would need some help to understand it. Could you give me an idea of what the test is that he is proposing?
Honestly? I'm still trying to make sense of it
(I didn't say it was a
good test, was just making you aware that it exists.)
By the way, don't you pay thirty bucks for that article. It has a
free version including all the supplementary material.
What Theobald seems to be doing is asking if the statement "these sequences evolved from a common ancestor in group X" explains their sequence similarity better than "these sequences are similar for unspecified other reason". ("These sequences" being 23 proteins shared by all cellular life forms.)
So he takes four species each of bacteria, archaea and eukaryotes, and seems to ask whether one common tree (or network) for all three groups fits the data better than two or three separate trees (or networks). Model comparison is by likelihood ratios, Bayes factors and something he calls the AIC that isn't quite the AIC. I don't really understand how you can get a meaningful test out of this once you allow horizontal gene transfer (which he eventually does)... Anyway, a unified tree/network for all three domains wins out over all alternatives by all criteria, and with quite large margins.
His critics in
Nature point out that he didn't truly take convergent evolution into account, to which he says that (1) there are currently no models of convergent molecular evolution but (2) convergence is unlikely to produce 50+% sequence identity independently given that
(1) analogous enzymes do the same job with no sequence OR structural similarity
(2) structural similarity doesn't require much sequence similarity anyway, certainly nowhere near as much as you find in these proteins
(Although if you can make that argument, I'm not sure what all the model comparison adds to it!)
Another critique comes from
Koonin and Wolf (2010 [this is also a free paper, including reviewers' comments!*]), who show that the mere fact of sequence similarity without an underlying evolutionary history can dupe these tests into voting for common ancestry.
(As far as I can see, Theobald's response to the first comment would also apply here - sure, sequence similarity can trick a likelihood ratio test, but why would there be
this much sequence similarity to begin with? - but again, if that is the root of his argument, all the tests seem kind of superfluous.)
*Which contain pretty scathing remarks about Theobald's paper itself... it's kind of funny to read, actually. Reviewer 1 definitely has style
