Hi all,
Just a note on this topic, which Rich may be able to help clarify if he
sees this. It can be shown that Rich's equation which uses log(abs(...)) is
exactly equivalent to the regular version of the Nee equation. However, for
an ultrametric tree, the ML of lambda will always be greater than
Hi all,
This (ML lambda mu) is a common belief that I think is mostly caused by the
fact that most implementations of Nee's equations constrain this.
library(diversitree)
set.seed(1)
trees - replicate(40, rcoal(100), simplify=FALSE)
get.ml.bd - function(phy, p=c(.1, 0))
Hi all,
Thanks for the quick replies, workarounds and bugfixes! Plenty for me to go on
with!
Simon
___
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Morning all,
Whenever I try to use ape's (v 3.0-2) birthdeath function e.g.
birthdeath(rcoal(sample(10:100, 1)))
I get the following error:
Error in while (foo(up) 0) up - up + inc :
missing value where TRUE/FALSE needed
The warnings all appear to be this sort of thing:
1: In log(1 - a)
Simon,
For reasons that I do not fully understand, the problem arises with the use
of a coalescent tree (which has a very branch length distribution than a
tree generated under a birthdeath process). maybe someone else has some
insight into why this causes birthdeath to fail.
if you run
If you want simple fitting of a constant-rate birth-death process to a
phylogeny, another option is in diversitree (see make.bd). Last time I checked,
bd in laser (and birthdeath in ape) used transformations that kept mu lambda,
but this is not the case for diversitree. The package is
Hi Simon,
This is fixed and you can find the new version on ape's SVN. Another
thing I fixed: it could happen that the calculation of the Hessian-based
SEs fails: the error is now caught and the MLEs are returned with SEs
set to NA (the profile-likelihood CIs are still computed of course).
