Multi-level Scoring Model Based On Item Node And Its Application In Multidimensional IRT

IRT is a new measurement theory which is to overcome the CCT. With the development of the IRT, measurement scientists also found that one-dimensional assumption of the IRT latent trait is not consistent with a number of psychological or educational tests. Therefore, the development MIRT is a very important task.In this paper, we firstly describes multidimensional item response theory and project scor-ing models, including multidimensional item response theory background knowledge, project scoring model of one-dimensional and multi-dimensional project scoring model.Secondly, the three different multi-level scoring models are derived based on the answers to three questions using node-based approach assuming that the probability of correct re-sponses was on the project node multidimensional two parameter logistic model. The first model is node-containing multi-level scoring model, multidimensional grade response mod-el, the latter two models are collectively referred to as nodes side-by-side multi-level scoring model, and then solve the scoring models of node-containing and nodes side-by-side items.Then, this paper studies the information function of project scoring model in MIRT. The information functions of multidimensional two parameter logistic model (compensation) and the multidimensional graded response model are obtained which is based on the information function extended from one-dimensional space to multi-dimensional space. We establish four conclusions by designing three experiments to analyses the influence of project parameters and the direction vector of information function experimental results.