CODEWITHSUNDEEP
rsummarylme4
Jonas
Jonas

Reputation: 1629

Summary of lme4 model in function (lmerTest)

If one has the following data

d = data.frame(out=rnorm(10), explain=rnorm(10), age=rnorm(10), sex=sample(c("M", "F"), size=10, replace=T), group=rep(c(1:5), 2))
f = as.formula("out ~ explain + age + sex + (1|group)")

and wants to fit a linear model with lme4 you can do

require(lme4)
require(lmerTest)
m  = lmer(f, d)
s  = summary(m)

That works, good.... But if the model is fitted within another function like

gglm = function(form, data){
    lm = lmer(form, data=data)
    return(lm)
}
m2 = gglm(f, d)
s2 = summary(m2)

I get an error. 

summary from lme4 is returned
some computational error has occurred in lmerTest

Apparently, this is because the fitting of the model was done with a object called data, which is not visible in the outer scope. So if I do data = d I get the same result as before. However, if I do 

data = data.frame(out=rnorm(10), explain=rnorm(10), age=rnorm(10), sex=sample(c("M", "F"), size=10, replace=T), group=rep(c(1:5), 2))

instead and get different data, the result of the summary is wrong.

This seems not to be the best way to do it and i think its easy to make mistakes. The normal lm and its according summary functiondon't have this problem. Isn't there a way to make the lmerTest summary less error prone?

Upvotes: 5

Views: 3223

Answers (1)

Dieter Menne
Dieter Menne

Reputation: 10215

The do.call trick works (but it is a workaround, for sure).

gglm = function(form, data){
  lm = do.call(lmer, list(formula=form, data=data))
  return(lm)
}
m2 = gglm(f, d)
s2 = summary(m2)

Upvotes: 2

Related Questions