基于Rasch模型的翻译测试效度研究

Validation of a Translation Exam Based on Rasch Measurement Model

本文运用多面Rasch模型，从考生、评分员和评分项三个方面对一次英语篇章翻译测试的效度进行了研究。结果表明：①考生能力具有显著差异。但是，部分考生的内部答题行为不一致：1．33％的考生未发挥出实际水平，7．34％的考生出现了超常发挥的现象，3％的考生则可能未译完；②评分员的严厉度存在显著差异，但他们的评分均具有良好的内部一致性；③评分项的难度存在显著差异，区分度也较合理。总乏，这次测试整体上具有良好的效度，但考生的表现值得进一步研究。

This study investigates the validity of an English text translation exam from three perspectives： examinees, raters, and rating items by implementing Many-facet Rasch Measurement Model. Results indicate that ① exami- nees are significantly different in ability, but some examinees display inconsistency in translating. 1.33% of the examinees do not bring their competence into full play when translating some sentences, 7.34% of the examinees translate some sentences unexpectedly well, and 3% of the examinees may not finish all the sentences; ②raters have significantly different severity, but their rating has good internal self-consistency; ③ rating items are significantly different in difficulty, and they differentiate the examinees well. In general, this study provides evidence for the good validity of the translation exam, but the performance of examinees needs further study.