How to get a model matrix from clmm objects?

Question

I want to estimate a multilevel ordered logistic model and afterwards access the model matrix. When running a simplified example from ?clmm:

library("ordinal")
mod1 <- clmm(SURENESS ~ PROD + (1|RESP), data = soup)
model.matrix(mod1)

I get the error message Error in eval(predvars, data, env) : object 'SURENESS' not found. From other packages I expected that setting parameters like model = TRUE the data going in are also exported to the estimated model, but here all relevant parameters seem to be set accordingly by default. Did I miss some parameters or elements from mod1 (I went through attributes(mod1) but did not find a model matrix.

Strangely if I set a random data.frame, it works:

set.seed(123)
df <- data.frame(y = factor(sample(c("A", "B", "C"), size = 1000, replace = TRUE), ordered = TRUE), 
             x = rnorm(1000), 
             id = factor(rep(1:10, each = 100)))

mod2 <- clmm(y ~ 1 + x + (1|id), data = df)
model.matrix(mod2)

So what's the difference between mod1 and mod2 and how do I get a model.matrix from mod1?

Answer 1

I do not think model.matrix(mod2) works for clmm objects. However, you can try to build a parallel model for the fixed effects part using functions like 'polr' and apply model.matrix() to the output object. The random-effects part can be fixed separately by using the clmm output.