Hello I'm working with mixed effects models using lmer() and have some problems to get all variance components of the model's random effects. I can get the variance of the random effect out of the summary and use it for further calculations, but not the variance component of the residual term. Could somebody help me with that problem? Thanks a lot! Below an example. Aline ## EXAMPLE #---------- require(lme4) ## Simulate data for the example set.seed(6) x1 <- runif(n=100, min=10, max=100) ## a continuos variable x2 <- runif(n=100, min=10, max=100) ## a continuos variable treat <- rep(letters[1:4], times=25) ## a fixed factor with 4 levels treat.effect <- 20*rep(1:4, times=25) group.label <- rep(LETTERS[1:5], each=20) ## the random effect group.effect <- 10*rep(1:5, each=20) ## there are 5 groups ## Response variable: y <- 2*x1 + (-5)*x2 + treat.effect + group.effect + rnorm(100) ## Dataframe d.ex <- data.frame(y, x1, x2, "Group"=group.label, treat) ## Apply model mod1 <- lmer(y~x1+x2+treat+x1:treat+ (1|Group), data=d.ex) output <- summary(mod1); output # ok, there is the variance component of the random effect group and the residual term ## Now I'd like to get the variance components of the random effect "Group" and of the residual term "Residual" in order # to do further calculations with these numbers output$varcor[1] ## reveals the variance of the random effect "Group" output$varcor[2] ## does not reveal the residual term! what other command do I need to use then?
Getting residual term out of lmer summary table
2 messages · aline.frank at wsl.ch, Ben Bolker
<aline.frank <at> wsl.ch> writes:
Hello
I'm working with mixed effects models using lmer() and have some problems to get all variance components of the model's random effects. I can get the variance of the random effect out of the summary and use it for further calculations, but not the variance component of the residual term. Could somebody help me with that problem? Thanks a lot! Below an example.
There's an r-sig-mixed-models at r-project.org mailing list that is specifically for mixed-models (especially lme4) questions.
Aline
## EXAMPLE #----------
require(lme4) ## Simulate data for the example set.seed(6) x1 <- runif(n=100, min=10, max=100) ## a continuos variable x2 <- runif(n=100, min=10, max=100) ## a continuos variable treat <- rep(letters[1:4], times=25) ## a fixed factor with 4 levels treat.effect <- 20*rep(1:4, times=25) group.label <- rep(LETTERS[1:5], each=20) ## the random effect group.effect <- 10*rep(1:5, each=20) ## there are 5 groups ## Response variable: y <- 2*x1 + (-5)*x2 + treat.effect + group.effect + rnorm(100) ## Dataframe d.ex <- data.frame(y, x1, x2, "Group"=group.label, treat) ## Apply model mod1 <- lmer(y~x1+x2+treat+x1:treat+ (1|Group), data=d.ex) output <- summary(mod1); output ## # ok, there is the variance component of the random effect group and the residual term
## Now I'd like to get the variance components of the random effect "Group"
and of the residual term
"Residual" in order
## to do further calculations with these numbers
output$varcor[1] ## reveals the variance of the random effect "Group"
output$varcor[2] ## does not reveal the residual term! what other command do
I need to use then?
VarCorr(mod1) ## prints standard dev by default in latest lme4
## print both var and std dev
print(VarCorr(mod1),comp=c("Variance","Std.Dev."))
unlist(VarCorr(mod1)) ## group variance (as raw number)
sigma(mod1) ## std. dev. of residual variance