error with predict function with lmer object-CodePudding

I'm trying to reproduce the plot from Roland's code with lmer instead of nlme model but I'm getting an error with the predict function:

library(nlme)
library(lme4)
library(ggplot2)

fm2 <- lmer(distance ~ age   Sex   (1|Subject), data = Orthodont)

newdat <- expand.grid(Sex=unique(Orthodont$Sex),
                  age=c(min(Orthodont$age),
                            max(Orthodont$age)))

p <- ggplot(Orthodont, aes(x=age, y=distance, colour=Sex))  
  geom_point(size=3)  
  geom_line(aes(y=predict(fm2), group=Subject, size="Subjects"))  
  geom_line(data=newdat, aes(y=predict(fm2, level=0, newdata=newdat), size="Population"))  
  scale_size_manual(name="Predictions", values=c("Subjects"=0.5, "Population"=3))  
  theme_bw(base_size=22)

This is the error I get:

Error in eval(predvars, data, env) : object 'Subject' not found
In addition: Warning message:
In predict.merMod(fm2, level = 0, newdata = newdat) :
  unused arguments ignored

CodePudding user response：

If you want to plot subject-level predictions, you would need to include Subject in your prediction frame:

pframe <- with(Orthodont,
  expand.grid(Sex=unique($Sex),
              age=range(age),
              Subject=unique(Subject))
)

Alternately you could include re.form = ~0 in your predict call: this is equivalent to the (ignored!) level=0 specification in your current code, which only works for lme ...