计分方式和题组效应对题组测验等值的影响:模型比较的视角  

Impact of the Format of Scoring and Degree of Testlet Effect on Test Equating for Testlet-Based Tests: the Perspective of Model Comparison

作　　者：李雨秦 

机构地区：[1]浙江师范大学教师教育学院,金华321004

出　　处：《心理技术与应用》2017年第6期334-340,352,共8页Psychology(Techniques and Applications)

摘　　要：采用模拟研究的思路,用项目反应理论(IRT)同时校准的方法对题组测验的项目参数进行等值;同时基于模型比较的视角,考查题组效应大小以及项目计分方式对等值结果的影响。研究结果表明:(1)不同模型在题组测验等值上的效果因计分方式和题组效应的不同而不同;(2)当题组效应较低时(0.5以下),等级反应模型(GRM)在区分度参数和难度参数上的等值效果均好于等级反应题组模型(GRTM),且不受计分方式的影响;(3)当题组效应较高时(0.5以上),模型的等值效果因计分方式而异,等级反应模型(GRM)在0/1计分项目的等值误差最小,等级反应题组模型(GRTM)则在多级计分项目的等值误差最小。A simulation study was conducted to investigate the impact of format of scoring and degree of testlet effect on test equating under polytomous IRT models. Both graded response model （GRM） and graded response testlet model （GRTM） were used to fit the equating results, meanwhile concurrent calibrate method was adopted to place item parameters that came from different tests on the same scale. Results showed that ： （ 1 ） The equating results under different IRT models were influenced by different formats of scoring and degree of testlet effect. （2） When testlet effect was small than 0. 5, GRM produced better equating results than GRTM both on discrimination and difficulty parameters regardless of formats of scoring. （3） When testlet effect was larger than 0. 5, the equating results under different IRT models depended upon the ways of scoring, GRM produced small equating errors than GRTM for di- chotomous items, while GRTM yielded small equating errors than GRM for polytomous items.

关 键 词：等值 IRT题组模型 项目参数等值 混合计分 

分 类 号：B841.7[哲学宗教—基础心理学]