Abstract: The Logistic, Normal Ogi ve, ARCTG, COS and LINEAR models in IRT have been compared by Monte Carlo Simulation approach. The result shows that the accuracies of parameter estimation of Logistic, Normal Ogive and ARCTG models are nearly equal; The accuracy of COS model is lower than the above three models; and, for the LINEAR Model, the accuracy is much lower than other four models in estimating item discri- mination than the other four models and is close to them in estimating item difficulty and examinees' ability.

Key words: NULL