2023 Volume 19 Issue 1 Pages 23-34
Multidimensional item response theory (MIRT) model is an extension of the unidimensional IRT model. Chen, Li and Zhang (2019) developed an estimation method called constrained joint maximum likelihood estimation (CJMLE) for the parameter estimation of the MIRT model. Moreover, they proved that its estimator of item parameter is asymptotically consistent under a rotation when the numbers of individuals and items simultaneously increase toward infinity. However, the rotation requires the true parameter. Hence, it cannot be applied in practice, when the true parameter values are unknown. In this paper, we investigate through simulation studies how the discrepancy between standardized true item parameter and rotated CJMLE solution changes under several rotation methods which seek the simple structure and are applicable in practice. The simulation results showed the non-decreasing discrepancy, and this could be attributed to the fact that standardized true item parameter does not have a simple structure.
