comment:
1)about the response time analysis, it utilized the multiple regression to demonstrate E1 responses are a result of intuitive reasoning by the observing E1 response are faster than E2. I think the response time can be analysis more. especially in CMR (figure 1), the theta C is consider that spending more time is necessary. Perhaps we can try to put a response time parameter in model for weighting the theta C. I mean we can let P(C) increase when the response time increases. As a preliminary attempt, we can try that theta plus a response time or multiplies a response time.
2)If we extend this research to CAT, the first difficulty of study is that item pool is too hard to expand. However, we still can try to find the information for this kind of multistage IRT. It should be useful for CAT or many other extended issue.
3) I think of applying this model in standard setting issue. When we want to set two cut point for three level examinee, such as professional, medial and poor. the judges set the higher cut point first and lower cut point second. It can be expected that setting lower cut point depend on the former higher cut point, because the lower point is always lower than former point. Thus, the probability of lower point is nested within the P(higher). The detail of this application should be further deliberated.