Hi All,
I am using pgls in the ¡®caper¡¯ package in R to test the
correlation between four explanatory variables and a response variable. Two of
the explanatory variables are continuous, the other two are discrete, and the
response variable is continuous. After running step-down model selection, my
final model only has a discrete variable as the explanatory.
During the model selection, I set all of the three branch
length transformation parameters (lambda, delta and kappa) to ¡°ML¡± for each
step down model. In each model, the range of each estimated parameter is:
lambda 0.887-0.896, delta 3, and kappa 1.021-1.347.
Here I have two confusions:
1. The optimum of delta is always set to the upper bound 3,
which looks strange to me. I learnt from somewhere that the delta parameter was
only for discrete traits but can¡¯t find the source of this contention. Is it
why it is always set to the upper bound? Should I still use this parameter?
2. It has been suggested that running multiple parameters at
the same time can make the branch length transformation hard to interpret. In
my situation, is it more appropriate to estimate only one of the parameters? If
so, which one should I pick?
Thanks,
Xu Han
Department of Biology, Queen's University,
Canada
[[alternative HTML version deleted]]
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
