摘要
根据PETS考试的特点,结合计算机自适应序列测试(CAST)的优势,研究者提出并设计了1-3-5三阶段和1-2-5-5四阶段两种PETS-CAST测试框架。为了检验两个测试框架的性能,研究者模拟生成了样本量分别为500、1000、3000、5000四个考生群体的能力值,然后,利用蒙特卡罗模拟方法,在试题参数已知的CAST框架上模拟生成考生的作答反应。研究结果表明,随着阶段的增加,自适应序列测试提供了更多的测验信息,能力估计的标准误逐渐减小,模拟能力与估计能力呈现出高相关。1-2-5-5四阶段测试对考生能力估计及分类决策的准确性更高,结果更为可靠。该模拟研究为PETS-CAST的具体实施在理论层面做了一些有益的尝试。
Based on consideration of the PETS characteristics and its future development, the authors start with design of two PETS-CAST configurations. One is a 1-3-5 three-stage test configuration; the other is a 1-2-5-5 four-stage test configuration. The simulation study is conducted to validate which one performs better. Real difficulty parameters of PETS items and the simulated ability parameters of the candidates are used to generate the original score matrix and the item modules are delivered to the candidates following two adaptive procedures set according to the path rules. The author simulates responses of 500, 1000, 3000 and 5000 respondents on 1-2-5-5 four-stage design and 1-3-5 three-stage design, and then compares the psychometrics indexes of the two CAST designs. Results show that the 1-2-5-5 four-stage CAST design is a little better than 1-3-5 three-stage CAST design. This simulation study provides a sound basis for the implementation of PETS-CAST.
出处
《中国考试》
2013年第1期16-22,共7页
journal of China Examinations
基金
全国教育科学"十一五"规划2009年度教育部重点课题"计算机辅助英语测试系统的研究与实践"(课题号:GFA097014)系列研究成果之一
