论文部分内容阅读
本研究采用概化理论和多层面Rasch模型,对26位CET-4作文评分人给30篇CET-4模拟作文的分数进行了分析。概化理论研究结果表明:考生能力是测试总变异的主要来源(77.7%),在当前采用一评的评分条件下,概化系数(0.81)可以接受。多层面Rasch模型分析结果表明:评分人严厉度差异显著,但自我一致性较好,评分人和考生偏差存在对高分和中分考生偏严、对低分考生偏松的问题,评分标准整体上能区分不同水平考生。
This study analyzed the scores of 26 CET-4 essayists on 30 CET-4 simulations using the generalized theory and the multi-level Rasch model. The result of generalized theory shows that Candidate ability is the main source of total variation (77.7%). Under the current condition of applying a comment, the generalization coefficient (0.81) is acceptable. The result of multi-level Rasch model analysis shows that there is significant difference between the severity of the scorers, but self-consistency is better. The scorers and the examinees have the bias to the high score and the middle score candidates. Can distinguish between different levels of candidates.