Hi, I have been trying glht() from multcomp package
and contrast() from contrast package to test a
contrast that I am interested in.
With the following simulated dataset (fixed effect
"type" with 3 levels (b, m, t), and random effect
"batch" of 4 levels, a randomized block design with
interaction), sometimes both glht() and contrast()
worked and gave nearly the same p values; other times,
glht()worked, but contrast() didn't and gave the error
message:
"Error in dimnames(X) <- list(dn[[1L]],
unlist(collabs, use.names = FALSE)) : length of
'dimnames' [2] not equal to array extent".
so apparently, the error message from contrast() is
data dependent. glht() always worked.
Here is the code, I just copy and paste in R
repeatedly:
batch<-as.factor(rep(1:4,3,each=20))
type<-as.factor(rep(c('b','m','t'),each=80))
y<-2*(type=='m')+4*(type=='t')+rnorm(240,100,20)
dat<-cbind(as.data.frame(y),type=type,batch=batch)
rm(batch,type,y)
library(nlme)
dat.lme<-lme(y~type, random=~1|batch/type, data=dat)
library(multcomp)
summary(glht(dat.lme, linfct = mcp(type
=c("b+t-2*m=0"))),test=Chisqtest())
library(contrast)
contrast(dat.lme, a=list(type=c('b','t')),
b=list(type='m'),type='average')
Does anybody have a clue?
Thanks,
John
____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
