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

Re: smallest pteros



----- Original Message -----
From: "Jaime A. Headden" <qilongia@yahoo.com>
Sent: Friday, October 22, 2004 12:53 AM

> Dave Peters (davidrpeters@earthlink.net) wrote:
>
> <I think I probably read that as "all branches" rather than "all trees."
> Mistake. What the heck is 'all trees anyway'? All possible trees? I'm
> confused.>

Yes, all possible trees.

>   Programs like MacClade and PAUP* produce more than one tree for a given
> dataset, and the more missing or "unknown" scored data ("-" or "?")

Warning!!! I don't know about MacClade, but PAUP* automatically interprets
"-" as a gap -- as a deletion in a molecular sequence (respectively as an
insertion in the other taxa). _Never_ put this symbol into a purely
morphological data matrix!!! You will get bogus results.

Which means that the tree published in the description of *Junggarsuchus*
_is_ bogus. OK, it's probably not, but the tree length is probably wrong, as
should be all support values and indices.

> the number of trees, which for say 150 characters and 70 taxa can
> be around 10,000 trees with a reasonably proportion of 10% missing data,

I won't look up the formula now, at quarter past one at night... I remember
how many strictly bifurcating unrooted trees there are for 9 taxa: 135,135.
For 10 there's well over a million.
        If you do an exhaustive search (as opposed to branch-and-bound or
heuristic), the program makes each and every of those trees, counts its
length and retains all shortest ones.

>   This is why I am skeptical of find only one single tree, because even
> with only 1% missing data in a 70 taxon, 175 character tree (making 12,250
> data fields, 1% of these as missing is going to result in over a 120
> "unknown" conditions for the whole, and that WILL influence the tree
> structure with every variation the program attempts to find for all taxa).

Oh yes.