IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

Tree reliability and the bootstrap

Joe Felsenstein joe at evolution.u.washington.edu
Mon Nov 1 19:55:35 EST 1993

Stephen Carr comments that

>	Bootstrapping is what you want. The bootstrap gives an indication 
>of how internally consistent your data are: if all of the informative
>(variable) sites give a consistent (or nearly so) phylogeny, any one resampling of
>the data is similar to any other, the same clades appear, and the bootstrap
>values are high. If on the other hand different portions of the data go
>in different directions, the reproducibility of any clade is low and the
>bootstrap values low as well. Bootstrap values are NOT statistical
>confidences, they are more like consistency checks - the higher the value,
>the greater the internal consistency of the data.

I beg to disagree.  Bootstrap P values are not confidence limits but
they do have straightforward statistical meaning.  If we pose the
question "if this group were not really there, what would be the
probability of seeing this much, or more, support for the group?",
it turns out that P is a conservative assessment of that.  It is just
as meaningful as a P value for a test of whether a treatment has
had an effect, in standard statistics.  For details of why this
interpretation works see the paper by me and Kishino in the most recent
Systematic Biology.

This interpretation is certainly different from the one in my 1985 paper,
but does fit into the use of P values in statistics much more neatly.  It
also is more than just a vague index of internal consistency.

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

More information about the Mol-evol mailing list

Send comments to us at biosci-help [At] net.bio.net