尽管多阶段测验(MST)在保持自适应测验优点的同时允许测验编制者按照一定的约束条件去建构每一个模块和题板,但建构测验时若因忽视某些潜在的因素而导致题目之间出现局部题目依赖性(LID)时,也会对MST测验结果带来一定的危害。为探究“LID对MST的危害”这一问题,本研究首先介绍了MST和LID等相关概念;然后通过模拟研究比较探讨该问题,结果表明LID的存在会影响被试能力估计的精度但仍为估计偏差较小,且该危害不限于某一特定的路由规则;之后为消除该危害,使用了题组反应模型作为MST施测过程中的分析模型,结果表明尽管该方法能够消除部分危害但效果有限。这一方面表明LID对MST中被试能力估计精度所带来的危害确实值得关注,另一方面也表明在今后关于如何消除MST中由LID造成危害的方法仍值得进一步探究的。 Although multistage testing (MST) allows test writers to construct each module and title board subject to certain constraints while maintaining the benefits of adaptive testing, ignoring some of the potential causes of construction testing can lead to problems In the event of a local problem-dependent (LID), MST test results will also bring some harm. In order to explore the issue of “the harm of LID to MST”, this study first introduces the related concepts of MST and LID, and then discusses the problem through simulation studies. The results show that the existence of LID will affect the accuracy of the test ability assessment The estimation error is still small, and the hazard is not limited to a specific routing rule. To eliminate the hazard, the group reaction model is used as the analysis model in the MST measurement process. The results show that although the method can eliminate partial hazards But the effect is limited. On the one hand, it shows that LID is harmful to the accuracy of the test ability estimation in MST. On the other hand, it shows that it is worth further exploring how to eliminate the harm caused by LID in MST in the future.