Dear list,
I am new to PCMs and confused about model assumptions using Pgls. I am
trying to determine the association between a discrete (categorial)
predictor and a continuous response variable across species using a block
of post burn-in trees. I have assessed model fit of the continuous response
using fitContinuous() assessing 4 models of evolution using AIC; Brownian
Motion provides the best fit and estimating lambda with phylosig() returns
a value =0.92 using the MCC tree. However, when I estimate lambda with the
gls model its 0.004. In fact a model with lambda fixed to 0 provides a much
better fit than the BM model or one where lambda is fixed to that estimated
using phylosig(). Somewhere in the process I've thoroughly confused myself
about whether there is phylogenetic independence of residuals error for
dependent variable in the gls model, and which is the appropriate model to
use (i.e., ols vs Pgls)?
Also, when trying to estimate phylogenetic signal for each gls model across
the block of trees my script crashes. From my understanding, this can
happen when lambda is very close to 0 or 1, and when I fix lambda to a
value the script runs fine. Considering the above, I was wondering if
anyone has any recommendations about how/whether to optimize lambda? If
needed, would be happy to send my script and data.
Thanks for all your help,
Jesse
PhD student
Boston University
[[alternative HTML version deleted]]
_______________________________________________
R-sig-phylo mailing list - R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
