Research on Identifying Central Rating in Net-based Composition Scoring Using Many-facet Rasch Model
YU Yunye · Zhongguo kaoshi · 2012
Researchers have found that in composition scoring process,some raters tended to assign scores, avoiding using the high or low end of the rating scale.This is the so-called central rating.This study tried to identify raters in net-based composition scoring using Many-Facet Rasch Model.The statistics were collected from MHK(LevelⅢ) composition rating which contained 15,194 examinees and 45 raters.This research estimated each examinee' s ability and expected score,and then by computing the residual' s standard deviation and coefficient with expected score,kurtosis of observed scores,infit statistics generated by the model,kurtosis of expected ratings and analyzing the result of expert-consensus scores,researchers successfully identified two suspected rater.Furthermore,this study suggested a series of standards to identify rating which could be applied in future net—based composition scoring.