The Parameter Estimation Of Multidimensional Item Response Theory Graded Response Model

Multidimensional item response theory (MIRT) is an important development of item response theory (IRT),The multidimensional grade response model can meet the needs of the application of such ratings. However, (MIRT) is still in the im-mature stage of development, the existing multidimensional grade model parameter estimation procedures have all kinds of defects.There are many parameter estimation Algorithms and parameter estimation procedures can be used to solve the problem of parameter estimation of he multi-dimensional grade response model. Such as N-R, E-M Algorithm, MCMC Al-gorithm,Metropolis-Hastings Robbins-Monro Algorithm(Cai Li,2010)、Noharm (Fraser,1998)、ConQuest (Wu, Adams and Wilson,1997)、Testfact、Mplus、IRTPRO、 flexMIRT and so on. but they have various defects and deficiencies.First,based on the DSY algorithm, this article developed the parameter esti-mation program of the MGRM and focused on three conditions:(1) estimate ability parameters when item parameters are known; (2) estimate item parameters when person parameters are known; (3) estimate item and person parameters simultane-ously when they are unknown. Then we compared the estimating results of DSY with MCMC algorithm using binary data. Finally, some further researches were recommended.