korbinian.strimmer at zoology.oxford.ac.uk
Fri Feb 25 15:52:22 EST 2000
> I have a large dataset (119 taxa) and using bootstrapping (NJ) or
> Quartet Puzzling (MP and ML) I get substantially different results.
> In general QP is much more conservative than bootstrapping, certainly
> in the deeper branches of the tree. Is this a general feature of QP
I think it is. QP really seems to be more conservative for many
data sets. The difference seems to lie in the fact that QP samples
the tree space directly (the p-step trees) whereas a convential
bootstrap analysis resamples the data. If the tree reconstruction
method is biased then bootstrap values will be at best misleading.
On the other hand, making a consensus tree of all trees sampled
from the tree space in the QP fashion may also lead to unnecessary
loss of information. In those case it is probably not a bad idea
to go back to the interim trees themselves (both for QP and NJ).
> methods, but I do not have the computational power at my disposal to
> perform 100 bootstraps using the ML or MP criteria.
Me neither :)
Korbinian Strimmer http://users.ox.ac.uk/~strimmer
Dept. of Zoology, Univ. of Oxford, South Parks Road, Oxford OX1 3PS
+44 1865 271272 (phone), -49 (fax), korbinian.strimmer at zoo.ox.ac.uk
More information about the Mol-evol