Some Studies On Item Parameter Estimation Methods In Logistic Model

In the last few years, Item Response Theory is a very active research area in educational and psychological measurement. Item Response Theory (IRT), which develops widely in various large-scale examinations, overcomes the limitations of Clas-sical Test Theory (CTT). There are many mathematical models of Item Response Theory. Logistic model, according to the number of parameters, can be divided in-to one-parameter, two-parameter and three-parameter three model. In the process of research and learning, we often involve the problem of estimating model param-eters, how to choose the better model is often be discussed. In view of this, this paper will introduce two estimation methods, marginal maximum likelihood estima-tion method(MMLE) and Bayesian estimation method, by which we can compare the advantages and disadvantages of two estimation methods with data stimulation,.First, this paper introduces the basic concepts and model of Item Response Theory, describes the difference between IRT and CTT in brief, analyses the item reflection model and the influence of parameters on the exams, and introduces the Logistic Model and characteristic curve. Secondly, we introduce the Newton-Raphson iterative algorithm and the EM algorithm. Next, we introduce the basic theory of marginal maximum likelihood estimation and Bayesian model estimation of the topic parameters, and obtain the item parameter estimation equation of two parameters model. At last, by data stimulation, we compare the advantages and disadvantages of the two estimation methods.