Peter J. Waddell
Department of Biology and Department of Statistics
University of South Carolina
Statistical Issues in Phylogenetics And Resolving the Inter-Ordinal
Relationships of Placental Mammals
The inferred evolutionary relationships of the 18 presently recognized
orders of placental mammals based on sequence data are settling into a
pattern very close to that of the tree proposed by Waddell, Okada and
Hasegawa (1999) on the cover of Systematic Biology. There remain
major unresolved statistical issues of how to test this, or any other tree,
for accuracy in both topology and divergence time. We explore the phylogeny
of mammals using the most comprehensive published data sets covering both
mtDNA and nuclear sequences at both the nucleotide and amino acid
sequences. There is good evidence that we cannot trust the current
statistics to give us a realistic view of the accuracy of groups. For
example, an exploration of recently proposed Empirical Bayesian methods
shows that they would appear to be far less robust than the bootstrap. It
is also now apparent that the bootstrap has on many occasions given far too
much support to incorrect mammal clades (for example that hedgehog is
sister to all other placentals). A statistical framework for testing clades
using SINE data is presented. Its application reveals significant support
for the tarsier/anthropoid clade (in contradiction to the most recent
sequence data), as well as the clades Cetruminantia and Whippomorpha (in
accord with most sequence data). Estimation of divergence times with
sequenced data is also very uncertain, requiring considerable faith in
untested assumptions. Divergence time estimates of placentals based on
improved methods of dealing with uncertainties in the data are compared
with previously estimated divergence times and highlight a number of
unresolved problems
Back to Colloquium Series