打通题库建设瓶颈的关键手段——基于Rasch模型的题目参数等值技术

Breaking through the Bottleneck of Item Banking——Linking Item Parameter Estimates to a Base Scale with Rasch Model

出于考试安全性的考虑,题目参数不等值的问题已经成了题库质量不高、作用不大,并直接损害考试信度、效度和公平性的一个突出瓶颈。文章以Rasch模型为例,结合WINSTEPS软件,对题目参数等值的意义、Rasch模型的特点、等值设计的方案,以及三种国际流行的题目参数等值方法,即同时校准方法、独立校准和链接相结合方法、固定参数校准方法,进行了详细解读。期待这些技术可以为我国的题库建设和考试工作带来积极意义。

Due to test security concerns, item parameter estimates are not linked to a base scale for most item banks in China, resulting issues in test reliability, validity and fairness. With Rasch model, the concurrent calibration, separate calibration with linking, and fixed item parameter calibration methods for linking item parameter estimates to a base scale are discussed. The equating design, as well as the WINSTEPS control file, is also introduced in detail. The methods discussed in this paper can benefit test development and item banking.