TEM4平行模拟测试信度及差异检验

Reliability and Variance Testing of TEM4 Parallel Simulation Tests

背景 :对于TEM4这样的全国性测试 ,要想稳定质量 ,经常性的评估是必不可少的。语言测试评估、有效性研究的一个重要组成部分就是信度或一致性研究。方法 :研究使用平行试卷 ,数据取自三次模拟考试和一次统考。统计检验从相关样本的特点出发 ,先进行信度统计 ,后采用差异分析。不仅检验了平行测试之间的一致性问题 ,还在有差异的情况下 ,对有差异的测试或题项进行定位。这种定位对以后的测试编制、预测及拼卷将起到参考作用。结论 :TEM4的信度适中稍低。

Background:This paper is an empirical testing of the reliability of TEM4. TEM4 is a national language test held annually, which entails close examination and constant evaluation,including reliability study. Reliability study is one of the indispensable procedures of validation of language tests. Methods: The data for this study have been collected from the parallel measures that subsume three TEM4 simulation tests using TEM4-99, TEM4-98 and TEM4-97 respectively and TEM4-2000, and involve the same group of students for all four tests. The statistical testing (SPSS 10.0) is, taking repeated measures as its departure, carried on around the four total means, from analyzing correlations to exploring differences. It not only examines consistencies between equivalent tests and also attempts to locate the variance in specific tests and test items, which will help make informed decisions in the development of TEM4 tests. Results and conclusions: The overall reliability of TEM4 is moderate, which may not be generalized across studies due to the defects of reliability coefficient per se and sample variability.