1. The pairwise comparisons are fully linked, that is, each alternative was comparied by all other alternatives directly. If, the pairwise comparisons are not fully linked, say, a is not compared with e direclty, will it affect the number of latent classes?
2. Will the idea applicable to ranking data where intransitive choice is not allowed, by definition.
3. No standard error iin Table 4, I think the SE may be very large.
4. What's saturated model in Table 3? No explaination. Does it mean each alternative has a preference order?
5. In the proposed model, the intransitivity was viewed as viewpoint shifts without any error. In practice, it is reasonble to assume that intransitivity was caused by error. Is it feasible to add an error term for zpq?