Unfolding IRT models have been developed in recent years, including the confirmatory multidimensional generalized graded unfolding model (CMGGUM; Wu & Wang, 2011). In this study, we implemented computerized adaptive testing algorithms based on the CMGGUM. The Fisher information was derived. Simulations were conducted to evaluate the performance of the algorithms, including the maximum a posteriori estimation for ability estimation, the maximum priority index (MPI) and D-, A-, T-, Eoptimality criteria for item selection, and a fixed-precision stopping rule. Results showed that all the four criteria achieved the predetermined precision level of .25, the bias in ability estimation was between -0.017 and 0.019, the mean square error was between 0.064 and 0.114, and the correlations between estimated and true latent traits was between .94 and .96. The mean test length for each dimension was similar, indicating that the MPI would facilitate content balance. Among the four optimality criteria, the E-optimality criterion was not recommended due to its longer tests and consuming time. The other three optimality criteria performed similarly. In addition, the differences of the Fisher information between unfolding and dominance IRT models were described.
|Publication status||Published - Jul 2013|