Thank you for all the tips here @drbenvincent. That makes sense - I’m comfortable with some simple re-parameterisations to please the sampler, but the Dirichlet approach was so far outside my intuition I couldn’t grasp it at all! It does make some more sense now and altering the min and max arguments makes things a bit clearer. I only tried quickly, but on the Kruschke data, fixing the cutpoints to the suggested yielded very similar estimates for the cutpoints and scale, but the mean was still slightly off.
I’m likely not going to be much use but I’m happy to contribute or test any ordinal regression resources. Would be amazing to see this woven into bambi one day in the future!
Thank you and happy holidays!