Ordinal probit response model (also translation from jags)

Modeling
6

Yes, you can do that. In the parameters{} block you’d set the number of cut-points “c_raw” to K - 3, and then in the transformed parameters{} block, you’d initialize the (complete) number of cut-points “c” to K - 1, and hard code the values of, say, the first two, and the remaining ones to the values of c_raw.

Something like:

parameters {
  ordered<lower = 0.5>[K-3] c_raw; // lower = 0.5 so parameters are restricted to be at least as large as the arbitrarily hard-coded cut-points
}

transformed parameters {
  ordered[K-1] c;
  
  c[1] = -0.5; // Your hard-coded values for the first, say, 2 cut-points. Arbitrarily choosing a value of 0.5
  c[2] = 0.5;

  for(i in 3:K-1) {
    c[i] = c_raw[i-2];
  }

}