A Simulative Study On The Combine Of The Classical Test Theory And The Item Response Theory

Classical Test Theory formed a set of statistic theories and operational methods through a century's development. Since its equations are simples and intelligible ,the theory has been applying with most of the educational and psychologic data and information of social science. With some limits itself, there have been some problems when it is used, though it have been a evidence in the field of social measurement, for example, parameter in the dependence of swatch, one error index a test, etc. However, getting over these limits, the item response theory became a developing and all-pervading modern measurement theory today, although its equations are complex, abstruse and strict qualification to real data, Of course, which block its expanse.In this dissertation, the author thinks that to combine the two theories can not only find some problems in the CTT's use, but also help to know the IRT's thought and method. This dissertation applies a simulation method, not the analysis according to real data, which are two factors:(1) can avoid the effect of the real data itself;(2) The simulation in compute is a method in common use in psychometrics, because in sometimes the pursuers need know parameters beforehand when comparing which methods is better.Firstly, this dissertation introduces the using models, in which the IRT model is thefoundation and others will be combined, i.e., Generalized Partial Credit Model.Secondly, the dissertation finds some problems, the appropriate condition and someconclusions further in CTTuse, using the programmers of MATLAB written bymyself, in which the computer produce the different population according torequest.In the end, in order to reference, the primary programmers and data attached.