Betty Schirrmeister <betty.schirrmeis...@googlemail.com> a écrit :
Hi
I am trying to reconstruct anestral character states with the *ace* function
of the *ape* package. I have a discrete character which among the taxa of my
phylogeny is present or not (1or 0).
When I use the "ER" model assuming equal transition rates between the two
states I get good results. But for confidence I wanted to conduct also a
Maximum Likelihood analysis with the "ARD" model assuming rates to vary and
finally compare the two results. *The problem is that some of my likelihoods
for the states of the ancestors are negative using the ARD model. *Does
somebody know why?
These negative values could arise when a model is not very good. In
your case, there are only two possible models (at least in this
framework), so that would suggest that the symmetric model (ER) is the
best one. Did you compare both models with anova()? Also maybe the
standard-errors of the asymmetric model (ARD) are quite large.
HTH
EP
Thanks a lot,
Betty
[[alternative HTML version deleted]]
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo