Plot Selected Random Effects Observations in Lattice

Question

I'm attempting to plot only certain observations from my random effects model (since the actual data set has a lot of observations).

Here is an example of the analysis:

# Load packages

library(lme4)
library(lattice)

# Load sleep data

data(sleepstudy) 

# Model

fit <- lmer(Reaction ~ Days + (1 + Days|Subject), data = sleepstudy)

# Plot random effects

dotplot(ranef(fit, condVar = T))

The plot looks like this:

What if I wanted to only plot the results for subjects 337, 310, 333, and 349?

I've tried saving the ranef() results, selecting the subjects of interest, and then building a plot from there but that wont work as I lose the error bars.

Answer 1

We need to manipulate the ranef.mer object.

library(lme4)
library(lattice)
data(sleepstudy) 
fit <- lmer(Reaction ~ Days + (1 + Days|Subject), data=sleepstudy)

First, we store it.

r.int <- ranef(fit, condVar=TRUE)

Second, we create a vector of the desired subset row numbers.

s <- c(337, 310, 333, 349)

Third, inside a lapply function we subset both in the list, the data.frame and importantly the attributes, where the variances are hidden in an array.

r.int <- lapply(r.int, function(x) {
  s2 <- which(rownames(x) %in% s)
  x <- x[s2, ]
  attributes(x)$postVar <- attributes(x)$postVar[, , s2]
  return(x)
})

Fourth we hack the required class label.

class(r.int) <- "ranef.mer"

Et voilà, we finally can plot our desired selection.

dotplot(r.int)

Yields