[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
Re: SICB Report
> In a message dated 1/18/99 10:39:26 AM EST, firstname.lastname@example.org writes:
> << Not correct. Different evolutonary conditions will most definitely
> impact the degree to which a strict parsimony analysis recovers the
> correct tree, or parts of it. This is something that we've seen in our
> simulations, and that Huelsenbeck and Hillis' work showed. >>
> My understanding of the situation is that under the best of circumstances
> (e.g., fossil record perfect), the analysis recovers the correct cladogram
> about 80-90% of the time. As circumstances worsen (more taxa missing, fewer
> specimens available, etc.), this percentage falls.
Which simulation study, specifically, are you basing your statements
on? And when you state that an analysis gets "the correct tree" a
certain percentage of the time - are you speaking of the entire tree, or
of parts thereof? This is a critical distinction.
Previously, I was willing
> to give analysis a 95% rating, hence my surprise to read that it's more like
> 85%. We >know< it's not 100%, so--what do >you< think the best-of-
> circumstances rating should be? And can I quote you?
Take a look at Huelsenbeck (1995), wherein some of his graphs definitely
show parsimony analyses (and other methods) recovering the correct tree
100% of the time under some circumstances. Differential branch length
seems to be the deciding factor. So you don't have to quote me - there
are peer-reviewed papers out there demonstrating it.
Huelsenbeck, J.P. 1995. Performance of phylogenetic methods in
simulation. Systematic Biology, 44:17-48.
This paper summarizes the preceding literature very well, but there are
more recent papers out there on simulated phylogenies, and I encourage
everyone to sit down with a thick stack of Syst Bio or any other
phylogenetics journal and read up. There are, indeed, situations in
which strict unweighted parsimony will not recover the correct tree, and
we don't expect 100% accuracy all the time - but to paraphrase
Huelsenbeck in his abstract, it does very well under most situations.
The most recent simulation is this one, which is very relevant to
Wiens, J.J. 1998. Does adding characters with missing data increase or
decrease phylogenetic accuracy? Systematic Biology, 47:625-640. And
read the whole paper - the abstract, in this case, is not an adequate
summary of what's inside.
(As I think I've said elsewhere, a quick reading of this paper, and some
others by this author, might lead to a confusion of resolving power with
accuracy. John's point is that with an increase in resolving power
comes an increase in the number of "incorrect" trees equally
parsimonious or likely with the true tree. This becomes a problem only
if we insist on having a fully-resolved tree for some sort of