Dimensionality analysis and modeling the structure of cognitive abilities by multidimensional item response theory based on local dependency indices

Dimensionality analysis or specify the number of latent traits that affect the data from the implementation of a scale or test and the study of local independency and monotonicity assumptions is of the important presuppositions to apply a item response theory model on a set of data. In this research, after a brief explanation of the local dependency indices and its models, In order to modeling the structure of cognitive abilities based on the data obtained from the implementation of the scale of cognitive abilities on a sample of 1105 students from the 9th grade students of Ahwaz selected by proportional stratified random sampling, after examining the hypothesis local independency through local dependency indices, monotonicity hypothesis and estimation of parameters of two one-dimensional and two-dimensional models and their comparison, The dimensionality of cognitive abilities was determined. The results showed that the dependency model observed between the data is perfectly locally dependent and Comparison of fitness indices of one-dimensional and two-dimensional model of Samejima graded response model showed that the two-dimensional model has better fit to the data than the one-dimensional model. In the following, using the concept of dimensionality factor space, a kind of comparison between the two classical test theories (CTT) and multidimensional item response theory (MIRT) was created and a kind of convergence in the results was obtained. The results showed that the fitness indices of classical theory alone were not able to characterize the structure of cognitive abilities, and methods based on item response theory in this field can be solved. The results of dimensionality analysis showed that the structure of cognitive abilities was two-dimensional; the first dimension of non-social cognition and the second dimension of social cognition were named and it was observed that these results converge with the results in the field of neuroscience.