In article <9r1hl0$o4a$1 at mercury.hgmp.mrc.ac.uk>, iotarho
<iotarhorho at hotmail.com> wrote:
> Good point that I should have specified -- these are rooted trees I'm
> working with. I've been using Molphy but I'm still reading up on how to get
> it to calculate likelihoods for all 3 (rooted) cases.
Making a rooted tree different from an unrooted tree requires either
specification of the states of the root (ancestor) or odd assumptions in
your evolutionary model such as non-reversible transition probabilities or
a molecular clock, neither of which I think you want to do. Do you?
Aren't these trees rooted with outgroups? If so, the outgroup is a 4th
taxon. If not, how do you root the tree? If by fiat, then lacking an
unusual evolutionary model there can be no difference in likelihoods among
the three trees.
> I've wondered if there has been any literature comparison of any distance
> methods versus ML specifically for a rooted, 3 taxa case? It seems that
> rather than comparing likelihood values that comparing branch lengths and SE
> measurements from NJ trees, for example, would be just as useful in
> determining relationships between three taxa. Any thoughts on this?
Again, differences among the 3 trees would require an unusual model of
evolution. UPGMA, for example, will do this for you, but do you want to
assume a clock?
> "John Harshman" <harshman.diespamdie at sjm.infi.net> wrote in message
> news:9qpurn$o8i$1 at mercury.hgmp.mrc.ac.uk...> > I can offer you very little help. But I do have a question: why is the
> > minimum of 4 taxa a problem? With 3 taxa there is only one tree, so there
> > can be no comparison of likelihoods among topologies. If you have a
> > program that works fine except for requiring 4 taxa, go with it.
> > And have you tried MOLPHY?
> > --
> > *Note the obvious spam-defeating modification
> > to my address if you reply by email.
*Note the obvious spam-defeating modification
to my address if you reply by email.