lavaan and Multiple Group Mediation

The lavaan tutorial on multiple groups has a good introduction to multiple groups, but doesn’t clearly specify one confusing aspect - estimating direct and indirect effects in multiple groups.

Specifically, say you have the following simple mediation model:

  'm ~ a*x
   y ~ b*m + c*x
   # indirect effect
   ab := a*b
   # total effect
   abc := ab + c'

The indirect effect is a*b and the total effect is ab + c. Running this with the group argument to sem() will in the best case return an error; though in some cases I see it run while producing identical estimates between groups.

Instead, you need to specify different parameters per model:

  'm ~ c(a1, a2)*x
   y ~ c(b1, b2)*m + c(c1, c2)*x
   # group 1
   ab1 := a1*b1
   abc1 := ab1 + c1
   # group 2
   ab2 := a2*b2
   abc2 := ab2 + c2'

This work is licensed under CC BY-NC 4.0