Hi James, The proportion of sites invariable should definately be
calculated for the entire data set as it has a very large sampling error.
The way PAUP* handles this for LogDet distances is to remove invariable
sites in proportion to base frequencies. If non-stationarity is a problem,
then base frequencies should probably be clalculated across constsant sites
rather than all sites (not the default in PAUP*). If non-stationarity is
not a problem, the point is moot. The person whose work most influenced
these ideas in both PAUP* and the Swofford et al chapter in Mol.Syst. 2nd
ed. is Peter Waddell. His e-mail is: p.j.waddell at massey.ac.nz.
regards, Jack
--
Jack Sullivan e-mail: sullivan at onyx.si.edu
Laboratory of Molecular Systematics Phone: (301) 238-3444
MSC, MRC-534 Fax: 301-238-3059
Smithsonian Institution
Washington, D.C. 20560
