【摘要】 等值是保证测验公平性的主要手段,主观性试题由于具有试题数量少、容易曝光、难度控制较难、评分受评分者因素影响大等特点,其等值问题一直是测量界一大难题。研究从实际问题和现实需求出发,使用了MHK口语的实测数据,采用非等组锚题的设计,将笔试客观题的听力理解部分作为锚题,尝试对4套口试试卷进行了IRT等值处理,并以随机等组的办法对其等值效果进行了检验。研究结果表明,把听力理解部分作为锚题进行口试试卷等值具有一定的可行性,因其中涉及许多问题,研究的结论还需进一步验证。更多还原

【Abstract】 Equating the test scores is crucial to the fairness and reliability of a test. Nowadays most test forms with multiple-choice(MC) items have been equated to make sure that the scores can be comparable in most large-scale assessments. Whereas tests with constructed-response(CR) items pose some challenge in the area of equating. The number of items used in CR tests is typically much smaller, moreover those items are easier to be exposed and the item difficulty is harder to be controlled. Also CR items tend to be difficult to score objectively and reliably. The problem discussed in this study comes from the real testing procedure, which meets the practical needs. Using data from one administration of MHK, a large-scale exam for testing Chinese minority‘s Chinese proficiency, this study investigates the use of MC items from the listening comprehension part as anchor items to equate CR items in the oral test of MHK in the context of the Item Response Theory methodology, using random group design to verify the equivalent accuracy. The results support the feasibility of the use of the proposed equating method. Since there are a lot of relevant practical issues in the equating procedure of CR items, the conclusion of this study needs further verification.更多还原