About the Bootstrap...

Dear all,

I have two simple questions about the bootstrap technique to estimate
confidence levels on edges and nodes.

First, it seems to me that authors have different definitions on what is
a *parametric* bootstrap. Is it: (i) a resampling procedure allowing to
assess confidence levels on an a priori phylogenetic hypothesis -- a
parametric topology -- given the available data, or (ii) a resampling
procedure without any a priori hypothesis but recquiring confidence
intervals (e.g., standard errors) linked to each descriptor (mean
values) for each analyzed taxon to be known in order to generate
pseudo-samples (e.g., by gaussian random drawing from N[mean; S.E.])? In
this second case, parametric bootstrap cannot be used with molecular
data and discrete morphological data (0/1), but can easilly be used with
biometrical and/or morphometrical data in association with distance
methods. If (ii) is not what is usually named a parametric bootstrap,
what it is? Is it known if confidence levels estimated with this second
resampling procedure have the same properties than more classic
nonparametric (1st order) bootstrap CL?

Second, since Efron's et al. (1996: PNAS, 93: 7085-7090) paper on second
order bootstrap confidence levels computation, does anyone here know if
algorithms and/or programs allowing to compute them in association with
distance methods (e.g., L.S. or N.J.)  have been published? I think the
last version of Phylip (3.6alpha3) does not allow to compute them -- it
should be an interesting improvment of this package!

Many thanks in advance,

Gilles Escarguel (CNRS/Univ. Lyon 1 - France).
Lab.: [EMAIL PROTECTED]