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

Re: Benton et al.'s Supertree (longer)

Rich Grenyer wrote-

> The problem comes when we 'admit' in a supertree
> (henceforth MRP) analysis that this difference in likelihood of
> correspondence to the true tree exists, because we suggest that some
> character sets a priori contain the most correct signal we can get, whilst
> other sets contain less correct signal. We then weight according to our
> ranking scheme if we wish.

When I write of a character set containing a "more correct signal", I'm
simply referring to the fact they have many more characters and taxa than
another "worse" character set, as I think we can all agree that adding
characters and taxa helps increase phylogenetic accuracy.  Another
consideration is that character sets I feel are "better" are generally more
modern, so have access to more data than older character sets.  Finally,
many of the trees included by Benton et al. are examples of the method you
detail below, where only characters that support the phylogeny are included.
A lot of these weren't even based on a phylogenetic analysis, the author(s)
just drew the cladogram and list which characters support which nodes (many
of which aren't supported by any listed characters).  You would hopefully
agree these "character sets" are "worse" than those used in larger actual
phylogenetic analyses.

> In morphological cladistics, this is equivalent
> to choosing and coding characters so that a particular topology will be
> favoured in the end - it's a cardinal sin, and invalidates parsimony as an
> optimality criterion because the final topology does not represent the
> minimal amount of feature (not character) change experienced by the group.

I completely agree here on the problem of only including characters that
support specific clades when performing a phylogenetic analysis.  A good
example is Sereno, as I wrote about here-
http://www.cmnh.org/fun/dinosaur-archive/2002Feb/msg00373.html .

> The ideal strict cladistic analysis will include the entire universe of
> possible informative features, coded as characters in an unbiased way, so
> that global parsimony sorts the winning signal from the losing
> with no a priori amplification,

Agreed.  This is what I'm trying to do with my (currently halted by college)
coelurosaur analysis.  Holtz's 2000 analysis is a great example.

> Obviously this is why strict cladists get into such frightful arguments -
> because no analysis is ever going to be like this, signal amplification
> always occurs, and the result is never objective (even if parsimony is an
> accurate reflection of reality in the first place).

The main factor hindering this is time, as one can always look at all
previous analyses and include all their characters.  It may take some time
to examine the characters yourself and code accurately (and others will
often disagree with your coding or character definitions), but you can make
a pretty successful attempt at testing all previously supported phylogenetic
hypotheses by including all their data and characters.  The problem of
course is that new characters and taxa are always being discovered, so
results never remain objective for long.

> But in MRP we're hugely
> privileged because we know that global parsimony is absolutely the correct
> criterion ('cos we chose it to be so), and that we have completely sampled
> the universe of possible characters ('cos there are only so many source
> nodes, and we've coded all of them) but that means we have to play by
> parsimony's rules.

But what makes a supertree better than an all-inclusive standard parsimony
analysis?  Supertrees that contain repeated characters in their included
datasets (like Benton et al.'s) commit the "cardinal sin"- they give
characters used more often greater weight.  However, an all-inclusive
standard parsimony analysis includes every character only once.  As the
various character sets used to construct a supertree are virtually
guaranteed to include many of the same characters, making a supertree that
includes all published characters without weighting problems is virtually
impossible.  Both supertrees and all-inclusive standard parsimony analyses
completely sample published characters and taxa, but the former do it at the
price of invalidating parsimony.

> In my opinion,
> far better to present an MRP tree as a strictly parsimonious consensus of
> source tree statements and if one's happy with that then maybe "an
> algorithmic consensus of phylogenetic opinion over a period of time".

I agree here, these are much better descriptions of what supertrees are than
the "beneficial combination of data that more accurately reflect
phylogenetic relationships" that many people probably think of them as.

Mickey Mortimer