PHYLIP versus PAUP?
joe at evolution.genetics.washington.edu
Thu Dec 4 13:12:21 EST 1997
In article <666i8v$7i9 at net.bio.net>, <newsmgr at merrimack.edu> wrote:
>1. Do I understand correctly that the weblike tree produced by the DRAWTREE=
>the PHYLIP is a unique feature of this package and is not available in othe=
>packages including PAUP*?
PAUP* not only has the ability to draw unrooted trees like this, but actually
is doing it using a new algorithm I told Dave about and urged him to use.
>... Actually an unrooted tree produced by DRAWTREE is
>very meaningful tree for me. On the other hand, a tree produced from the sa=
>input file by DRAWGRAM looks very different for me. For the DRAWGRAM to gi=
>me a tree topology which I can interpret the same way as in the DRAWTREE tr=
>have to root it at the midpoint.
This is a matter of what your eyeball tells you. If you unroot the tree
that is in DRAWGRAM, you get the tree in DRAWTREE. If you assume that the
root in DRAWGRAM means something, when it doesn't, you will be kidding
>2. For the bootstrap analysis what is the bootstrap cutoff value for=20
>statistically significant and not significant clustering? I saw some author=
>are using 70%. Is it commonly accepted figure? Exactly what level of
>confidence 70% bootstrap value provides?
It used to be 95%. But even then, we realized that if you scan a number of
groups and pick the one that has highest bootstrap value, you have a
"multiple tests problem" and it is not as significant as it seems. More
recent work by Zharkikh and Li and by Hillis and Bull have verified that
in fact, the P value underestimates the true P. Hillis and Bull's simulations
suggested that 70% for a previously-chosen group is good enough, but the
matter needs more work to know how general this is.
Joe Felsenstein joe at genetics.washington.edu
Dept. of Genetics, Univ. of Washington, Box 357360, Seattle, WA 98195-7360 USA
More information about the Mol-evol