That sounds interesting, though I’m not sure I have the custom likelihood chops for that either ;)
A related question I had was how to set the priors for these ordinal-type variables.
I understand that the output has multiple intercepts corresponding to the thresholds. I tried priors for “Intercept” generally at Normal(0,1) and Normal(0,5) – I guess this is in the logit scale as for other brms models.
However, this means that the left- and right-most intercepts would be pretty far away from this prior, right? Should I specify a different prior for each intercept, as I’ve seen here?