The paper developed a subdimensions Rasch model to consider the local dependence among the items of the same subtests. It was declared that the proposed model could yield more precise measurement for the item and person parameters.
Questions:
1. How to explain the correlations between the testlets if the correlations are high?
2. It was said that the subdimensional models have a restriction of equal variance between subdimensions. If the variance is not equal, the subdimensional model could not model the LID to full extend. And if the variance is equal, the subdimensional model yields results almost equivalent to those the unrestricted multidimensional model. But how can know whehter the variance is equal?