'drop_unused_levels' argument for interactions

Interfaces brms
1

Hi, I was wondering, if the ‘drop_unused_levels’ argument only works for single terms, but not for an interaction between two factors. In the following example, I have assumed, that the interaction term “var1a:var2z” is dropped (as in rstanarm).

data <- data.frame(y=c(1,2,3,4,5),
                   var1=c("a","a","b","b","b"),
                   var2=c("x","y","x","y","z"))

fit <- brms::brm("y~-1+var1:var2", data=data, drop_unused_levels = T)

Win 10
R 4.3.0
brms 2.20.3

1 Like
2

I have the same question - is it even possible to drop only interaction terms?