Error using mediation package with lme4 model: mediator model is not yet implemented

2.5k views Asked by At

I have a data.frame, df:

> str(df_ss)
'data.frame':   571 obs. of  4 variables:
 $ final_grade  : num  0.733 0.187 0.502 0.194 0.293 ...
 $ time_spent   : num  -0.2 -0.326 -0.709 -0.168 -0.254 ...
 $ gender_female: num  1 0 1 0 0 0 1 1 1 1 ...
 $ course_ID    : Factor w/ 26 levels "1","2","3","4",..: 14 18 13 21 24 15 3 24 9 13 ...

I am trying to see how time_spent moderates the relationship between gender_female and final_grade. I'm specifying a random effect for course_ID.

The models I specified using are as follows:

med.fit <- lme4::lmer(time_spent ~ gender_female + (1|course_ID), data = df)
out.fit <- lme4::lmer(final_grade ~ time_spent + gender_female + (1|course_ID), data = df_ss)

Those seemed to work fine.

Following an example using the lme4 package in a vignette for the mediation package, I specified this mediation model:

library(mediation)
med.out <- mediate(med.fit, out.fit, treat = "gender_female", mediator = "time_spent", dropobs = T)

This led to this error output: Error in mediate(med.fit, out.fit, treat = "gender_female", mediator = "time_spent",: mediator model is not yet implemented.

Per this mailing list question (and answer), I checked that:

  1. inherits(mediatorModel, "merMod") returned TRUE and
  2. getCall(mediatorModel)[[1]] returned lme4::lmer
1

There are 1 answers

0
Weihuang Wong On BEST ANSWER

Instead of lme4::lmer, you might try loading lme4 using library(lme4), then just call lmer. Looking at the mediate code shows that the error-handling checks are looking for an exact match for lmer i.e. getCall(model.m)[[1]] == "lmer".