Tree reliability

Joe Felsenstein joe at evolution.u.washington.edu
Mon Nov 1 19:45:17 EST 1993


In article <1993Oct28.173222.27104 at gserv1.dl.ac.uk> gertjan at nl.kun.sci (gertjan) writes:
>I have a set of orthologous sequences of approximately 150 amino acids
>from ten different animal species. The sequences don't differ very
>much, most of them in about ten positions.
>I have made maximum parsimony and neighbor-joining trees with these
>sequences, which are looking quite nicely, but how can I say
>anything about the reliability of these trees?
>Bootstrapping would clearly not be of much use, since in the
>resampling process often all significant sites will be removed.
>Or do I just have to accept that trees based on such small data sets
>will never be very reliable?

Well, bootstrapping is best described as making a worst-case analysis,
and if you accept that as your viewpoint, you can't do better.  If all
ten positions are completely compatible, they may help convince you
that they are all "clean" in which case when all ten are omitted you
need not be upset, for then you at least have nothing contradicting the
answer.

The justification for bootstrapping is asymptotic and works best for
large data sets, so with only ten sites one will to some extent find
angels dancing on the heads of pins.

-----
Joe Felsenstein, Dept. of Genetics, Univ. of Washington, Seattle, WA 98195
 Internet:         joe at genetics.washington.edu     (IP No. 128.95.12.41)
 Bitnet/EARN:      felsenst at uwavm



More information about the Mol-evol mailing list