Research Of Parameter Estimation Based On Cos Method

How to choose a parameter estimation method reasonably is one important question in application of Item Response Theory.There were some parameter estimation algorithms have been researched such asχ2/EM, SQRT/EM,and rectified MIDI algorithm.The paper researched Cos method.It's a new method which has better efficiency in test equating.In this paper,its application in parameter estimation has been researched and some conclusions have been concluded.The paper include the following five parts:IRT and parameter estimation were introduced in the first and second part,separately.The parameter estimation based on Cos method was researched in the third part;and some conclusions were concluded through computer simulation in the fourth part.The fifth part was summarized and final part is the reference which been applied in the paper.Cos method is introduced and illustrated through computer simulation and some conclusions were also outlined:①The accuracy of parameter estimation is explicitly improved following the increase of examinees when the items'quantity is fixed;②Parameters'ABS and RMSD are very small when the examinees'quantity beyond 5000,and the credibility of the estimation is relatively high;③Estimation's decision is also explicitly improved following the increase of items'quantity when the examinees'quantity is fixed;④Estimation error is relatively high and the method's robust is relatively poor when the items'quantity less than 11;⑤The estimation's accuracy will be relatively high if the examinees'quantity is too small;⑥The estimation's accuracy will be relatively high as the increasing of K,and the time of computation will be longer.The paper's creative point is Cos method .It is a new parameter estimation method which have been researched in the paper.