With the aim of removing human errors and providing critical feedback and suggestions for improvement, considerable research has be done on computer-based automated essay-scoring systems. Examples of these include e-rater, PEG, IEA, IntelliMetric, and BETSY. This paper summarizes how these systems work in an attempt to comprehend their features. They are also compared. An automated Japanese essay-scoring system named Jess is introduced, including our analysis of its performance. Lastly, difficulties caused by its treatment of Japanese passages and related problems are discussed.
In this study, we proposed an estimation method for equating coefficients of the continuous response model based on the common examinees design. This method applies the EM algorithm according to the Shojima (2003) method, but does not require numerical approximation in E-step or numerical iteration in M-step. We also presented a general framework for when data were missing at random and confirmed the unbiasedness of the proposed estimator using simulation studies. Finally, we illustrated an example of numerical analysis on a real data.
In this research note, we examined the degree of difference/concordance among the response rates and the differences between the response rates due to the effect of weighting on the response data acquired from two surveys, the Health and Culture Survey and the East Asian Value Survey. The differences between the response rates, as well as the scale values between the two surveys, were relatively small and the structures of the response patterns were relatively similar. These findings indicate that the reliability of the survey results is relatively high. The difference between the original and the weighted Korean survey proved to be relatively minor (maximum difference of 3.5%), and the effect of weighting on the response data proved negligible. It has been recognized that comparison of pattern structures in groups of multiple questions, rather than comparison of response rates for each question, is important when seeking cross-national comparability.