Thanks very much Simon, that is super helpful.
Best,
Chris
On 25/3/20, 9:47 am, "Simon Wood" wrote:
Hi Chris,
It's kind of a documentation glitch, a node is not supposed to be listed
as its own neighbour (it causes the diagonal entries in the penalty
matrix to be over-w
Hi Chris,
It's kind of a documentation glitch, a node is not supposed to be listed
as its own neighbour (it causes the diagonal entries in the penalty
matrix to be over-written by the wrong value). i.e. the neighbour list
should be.
NB <- list()
NB$'East Timor' <- c(2,15)
NB$Austral
Hi David,
Thanks for the comments. I am running the analysis on a mac OS 10.12.6, using
R R 3.5.3 GUI 1.70 El Capitan build (7632), and mgcv 1.8-31.
I am aware of the procedure of using data = "xx" in a call to gam. I am having
a strange issue locally, in that gam does not see to be able to
On 3/18/20 12:44 AM, Wilcox, Chris (O&A, Hobart) wrote:
Hi all,
I am trying to fit a model with a markov random field smooth in mgcv. I am having some trouble with getting it to run, and in particular I am getting the message
Error in initial.sp(w * x, S, off) : S[[1]] ma
Hi all,
I am trying to fit a model with a markov random field smooth in mgcv. I am
having some trouble with getting it to run, and in particular I am getting the
message
Error in initial.sp(w * x, S, off) : S[[1]] matrix is not +ve definite.
After reading everything I
5 matches
Mail list logo
