31 Calibration of response data using MIRT models with simple and mixed structure (Present by Jacob)

xiaoxue‘s review

xiaoxue‘s review

by KUANG XIAOXUE -
Number of replies: 0

Calibration of Response Data Using MIRT Models With Simple and Mixed Structures

Jinming Zhang

The principal of JMLE and MMLE approach were introduced. U nder the unidimensional assumption,

unidimensional and multidimensional approaches have the same parameter estimated. While MMLEs are

different unless the correlation coefficients between underlying dimensions of simple structure are known to be

zero. So Simulated study1 were used to compare the accuracy of item parameter estimates obtained from the unidimensional and multidimensional approaches when the MMLE method was applied.

The estimation program ASSEST was selected to estimate item parameters in this study.

The results show that when the correlation between subscales is zero, these two approaches are basically the same. When the number of items is small, the multidimensional approach provides relatively more accurate estimates of item parameters; otherwise, the unidimensional approach prevails.

Simulation2 is used to investigate the impact of the violation of the simple structure assumption. The results

demonstrate that inaccurate estimation results may be obtained if an Mixed structure is incorrectly specified as

an simple structure.

Comments :

1 if the test contains more than 2 dimensions, what will happen to the results ?

2 if the correlation between the dimensions are higher than 0.9, then there will be higher latent variable, it will go back to testlet or two-tier model, then we do not need to use this model?