The estimators of IRT commonly treat the item parameters as “true” parameters and ignore the potential uncertainty. It will, definitely, result in incorrect estimates of the standard errors of measurement (SEMs). The authors argued that the issues become serious when the sample size is small.
The study proposed a multiple imputation-based approach, tending to show that it can obtain correct estimates of individual scale scores, as well as the SEMs. A simulation study was used to investigate the performance of the proposed approach.
Questions & Comments:
1. When the number of items and dimension increase, the imputation-based approach will be tedious and hard to implement.
2. It is interesting to simulate the consequence for SEMs when the uncertainty of items was ignore.