Hi Liam,
Thanks for checking on that, but Gsize needs to be ordinal, so the code I used
to convert it in R would be:
NewMod56$GsizeF<-ordered(NewMod56$GsizeF, levels=1:4)
I apologize I forgot to include this before.
-Original Message-
From: Liam J. Revell
Sent: Wednesday, November 18,
Hi Nate,
Thanks – I understand your point. But when I use Gsize as continuous they run
fine, but with Gsize as an ordinal factor, they are singular. The numbers used
in Gsize are identical in each case, but the change from continuous to a factor
is what triggers the error. So I don’t think it’s
Hi there Ted:
Yeah I wouldn’t expect this to be an issue with the VertLife Mammalia trees, as
all branch lengths in the whole tree are >0 (its calibrated to positive time).
More likely, I’d guess that the "computationally singular” error is telling you
that you lack variation in your trait
Thanks Liam for the quick response!
I just checked this with NewMod56dnaTree$edge.length and all of these values
are >0. The smallest edge length is 0.12 and the largest is 38.6. This tree is
from the Upham et al 2019 DNA-based mammal supertree - it's just been pruned
down to our dataset. So
Hi Ted.
My best guess is that some trees have zero-length terminal edges
resulting in implied correlations between related species of 1. Could
that be the case with your trees? Note that a good solution to this
problem *is not* to add a small amount of edge length to the tips
causing the
Hello all,
I'm running pgls analyses using caper and have run into an issue with getting
computationally singular errors. I'm trying to run a continuous dependent
variable (HeadShades) against an ordinal factor (GsizeF). I get the following
error when I run this:
Error in solve.default(xVix,
