reweighting characters, few and many

[Jan Bosselaers]

Thomas,

> Already a few months ago, an anonymous list member told me that cladistics
> functions like that: Figure out a tree that you personally like! In the next
> step choose a suitable program and algorithm which generate exactly this tree.
>
> Yesterday, at night, I made my first exciting experiences of this kind.
>
> In PAUP 4.0, I ran a matrix which gave 423 most parsimonious trees. As the
> consensus tree was highly unresolved, I ran the command "reweight"

Which index did you use? Did you use maximum, minimum or mean fit?

> which gave,
> after a new search, 21 parsimonious trees, and a nice consensus tree, but one
> that I personally dislike.
>
> So, in the next step I widened up slightly, but only slightly, the upper bound
> of the trees to be kept and searched the unweighted characters again.
> Excitingly, this enlargened the tree file from 423 trees to approximately 25000
> trees. Now, I reweighted those 25000 trees, and run parsimony search again.
> I got only three trees, and Imagine!!!, the consensus tree corresponds exactly
> to the tree, which is my person favourite!!!

How many iterations did you need to arrive at the three trees? I am not
sure, but I have a gut feeling that it is theoretically unsound to start
a successive weighting procedure from a set of suboptimal trees. I did
some experiments with a small dataset and obtained totally unreliable
results, in one case iterations swingin back and forth between two
results without ever stabilising.

Best wishes,

Jan
--------------------------------------------------------------
Jan Bosselaers
"Dochterland", R. novarumlaan 2
B-2340 Beerse, Belgium                       tel 32-14-615896
home: dochterland at village.uunet.be           fax 32-14-610306
work: jbossela at janbe.jnj.com
web: http://gallery.uunet.be/Dochterland/
"Aging is a disease, and I hate it" Christian Barnard