This paper introduce two approaches for estimating the accuracy of classification. One is Lee approach based on total score and the other is Rudner approach based on latent trait of IRT. The different IRT models, test length, and cut score location are compared in this study. The conditional classification accuracy is unrealistic in practice because the true category the examinee belong in is needed. Therefore the Lee approach and Rudner approach are used to estimate the classification accuracy. The true accuracy which could be obtained in simulation study is the base line to evaluate the effects of different manipulated conditions.
1) Whether the classification accuracy could instead of the standard error of classification? I think these two kind of indices are similar with each other.