Hi Dan,
This is indeed an issue. I had the opportunity a few months ago to get
the outputs from r8s and chronopl on the same data: this showed some
difficulties with r8s, namely the starting and final values of the PL
were identical.
I'm currently working on chronopl because it fails to find decent
starting values when several nodes are defined within intervals (it's
looping endlessly). I've more or less solved this problem and this
should be fixed in ape 3.0.
I admit that chronopl has some weaknesses. I can see three explanations
(all non-exclusive):
1) The implemantation in chronopl is flawed in which case it can be fixed.
2) The model of correlated rates is not appropriate in most cases. I
remember to have seen a paper stating that the relaxed clock model
(without correlation of rates) better fits real data.
3) The problem of optimizing a function of so many parameters is too
difficult for the currently used algorithm in which case a Bayesian
approach might be more appropriate, or using a better optimization
algorithm.
Cheers,
Emmanuel
Dan Rabosky wrote on 02/02/2012 22:55:
Hi Brian-
It has been awhile since I played with chronopl(), but at the time, I convinced
myself that results were substantially different from penalized likelihood as
implemented in r8s. I haven't explored this exhaustively, but at the time, I
was very much convinced that the results of chronopl were potentially
problematic. Other folks have mentioned to me that they've observed similar
differences between r8s and chronopl and that they've recovered bizarre
smoothed trees from chronopl. Perhaps someone with more recent chronopl
experience can weigh in on this...
~Dan
On Feb 2, 2012, at 6:06 AM, Brian O'Meara wrote:
Ape can do this, function chronopl. In general, one good way to find out
which package can do something is by looking at the task view for
phylogenetics: http://cran.r-project.org/web/views/Phylogenetics.html .
This will also allow you to install all the phylogenetics packages (at
least those on CRAN) with just a couple of commands. An updated version
should be coming out later today (and if someone notices other things
missing, please let me know).
Brian
_______________________________________
Brian O'Meara
Assistant Professor
Dept. of Ecology& Evolutionary Biology
U. of Tennessee, Knoxville
http://www.brianomeara.info
Students wanted: Applications due Dec. 15, annually
Postdoc collaborators wanted: Check NIMBioS' website
Funding wanted: Want to collaborate on a grant?
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
--
Emmanuel Paradis
IRD, Jakarta, Indonesia
http://ape.mpl.ird.fr/
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
