1. Given the good idea that an examinee may answer those easier items first and leave the harder items till a later period in an time-limit test, from the proposed LHL-2PL (Equation 2), it states that the probability of answer an item correctly is a product of IRT model and a speed term. The speed term which is in a range of exp(-lambda*(bj-taup)) and 1, decides whether the item would be directly answered or retained. However, I'm curious how to retain the items to a later period through the speed term? Would the retained items be answered later? If yes, will the probability different from the one which is defined in Equation 2?
2. Are the RMSE in Table 1 and 2 acceptable?
3. It is reasonable that an additional speeded factor for the harder items. Thus, it is interesting to compare the performance of MIRT and the proposed LHL-2PL in the empirical study.
3. Imaging the responding procedure, maybe it is likely that different person has different speeded rate and thus a subscribe p can be incorporated into the parameter lambda. It is desirable for future study.