Dear all,
I am quite new at R and have a question about using lme with crossed random
factors. I followed the instructions of Pinheiro & Bates, but that did not
work because of the non grouping of my data. Reading prior threads (
http://www.mail-archive.com/r-help@stat.math.ethz.ch/msg10849.html), I found
a solution to deal with non grouped data and crossed random factors in lme,
by defining a grouping with one level and using it in the random part of the
lme formula in the following way:
one <-rep(1,length(y))
random=list(a=~1,one=~b)
but with the comment, that I get "DF calculations, none of which are
correct
in the completely balanced case".
Does that mean, that calculations in my summary table (e. g. F- and
p-values) might be wrong? Has there arised another solution for crossed
random factors in lme with non grouped data in the meantime?
What will change in the model if I would use the formula vice versa or for
both random factors:
random=list(one=~a,b=~1)
random=list(one=~a,one=~b)
To which random factor do I have to add "one", dealing with three
random
factors?
Is it possible to use crossed and nested random factors together in one
formula?
e.g.: random=list(a/c=~1, one=~b)
I can't use lmer because I also have to deal with heterogeneity using the
varIdent function (what works brilliant).
I hope that I did not miss a solution to my problem in another thread or
that my questions are too naive!
Thanks in advance for your help and comments!
Regards
Björn Klatt
[[alternative HTML version deleted]]
