加速作答行为的混合项目反应树模型研究  

The Mixture Item Response Tree Model for Speededness

作　　者：黄庆 王倩 闻家君[4] 涂冬波[2] HUANG Qing;WANG Qian;WEN Jiajun;TU Dongbo(School of Foreign Languages,Quzhou University,Quzhou Zhejiang 324000,China;School of Psychology,Jiangxi Normal University,Nanchang Jiangxi 330022,China;Career Development Center,Suzhou Science&Technology Town Foreign Language School,Suzhou Jiangsu 215100,China;The High School Affiliated of Jiangxi Normal University,Nanchang Jiangxi 330046,China)

机构地区：[1]衢州学院外国语学院,浙江衢州324000 [2]江西师范大学心理学院,江西南昌330022 [3]苏州科技城外国语学校生涯发展中心,江苏苏州215100 [4]江西师范大学附属中学,江西南昌330046

出　　处：《江西师范大学学报(自然科学版)》2024年第5期449-458,共10页Journal of Jiangxi Normal University(Natural Science Edition)

基　　金：国家自然科学基金(32160203)资助项目.

摘　　要：在心理与教育测验中,通常假设被试会全力以赴完成所有测验项目.但在动机、时间压力等因素的影响下,被试在测验中不可避免地会出现各类异常反应行为.加速作答是一种常见的异常反应,广泛存在于各类测验中.为了更好地在实践中处理含有异常反应的加速作答数据以及提升对被试能力估计的精度,该文在项目反应树(item response trees,IRTrees)模型框架下,首次尝试对2种常见的加速作答行为(即随机猜测和表现下降)进行混合建模,开发了一个考虑加速作答的混合项目反应树模型(mixture IRTree model).研究结果表明:混合IRTree模型在各种实验条件下均有良好的参数返真性;同时实证数据分析发现:混合IRTree模型比传统IRT模型能更好地拟合真实数据,且更能反映被试真实的能力水平.In psychological and educational measurement,it is commonly assume that test-takers attempt all test items with full effort.However,under the influence of factors such as motivation and time pressure,the test-takers will inevitably exhibit various aberrant response behaviors in the test.Test speededness is one of the most common aberrant response behaviors that is widely present in various tests.In order to better handle the speeding answering data with aberrant response and improve the accuracy of estimating the test-takers'ability,under the framework of item response trees(IRTrees)model,two common speededness behaviors,namely random guessing and performance decline,are simultaneously modeled for mixture modeling for the first time in this paper,and the mixture IRTree model considering speededness is developed.The research results show that the mixture IRTree model has good parameter recovery in various experimental conditions.The empirical data analysis finds that the mixture IRTree model can better fit the real data compared with the traditional IRT model,and can more reflect the true ability level of the test-takers.

关 键 词：项目反应理论 树模型 混合模型 加速作答 随机猜测 表现下降 

分 类 号：B841[哲学宗教—基础心理学]