论文部分内容阅读
信度是任何测试结果有效的必要条件。为探究CET-6写作评分的信度表现,本研究通过使用概化理论和多层面Rasch模型,对10名CET-6评分员对100份CET-6实考作文的评分结果进行了分析。概化理论的分析发现,评分员侧面以及包含评分员与考生间交互作用的残差的方差分量在总方差中占有一定的比重。而多层面Rasch模型的分析则发现评分员在严厉度上的确存在较大的差异;而且评分员与考生间的显著偏差交互也呈现出对较高能力的考生偏严,而对较差能力考生偏松的趋势。研究也表明概化理论和多层面Rasch模型具有良好的互补性,能对测试信度做出点面结合的丰富说明。
Reliability is a necessary condition for the validity of any test result. To explore the reliability of CET-6 writing scores, this study analyzed the scores of 100 CET-6 test essays by using 10 generalized theory and multi-level Rasch models. The analysis of the generalized theory found that the variance component of the residuals on the side of the scoring staff and on the interaction between the scoring staff and the examiner occupies a certain proportion of the total variance. However, the analysis of multi-level Rasch model shows that there is a big difference between the severity of the scorers and that the significant deviation between the scorers and the examinees also shows the bias towards the candidates of the higher ability. However, for the poor ability candidates, Partial loose trend. The research also shows that the generalized theory and the multi-level Rasch model have good complementarity and can make a rich description of the test reliability.