双因子模型下CAT测验优化设计及其效果验证

The Optimization of Testing Design for CAT with Bifactor Model and Its Application

在2种传统的BCAT测验设计的基础上,提出了4种新的BCAT测验设计,并采用国际上通用的Monte Carlo模拟实验的方式,从被试能力参数估计精度、题库使用的曝光率及测验的效率等3大指标来验证新开发的4种BCAT测验设计,再与传统的BCAT测验设计进行比较,以验证该文提出的4种新的BCAT测验设计的科学性、效果及优势.最后,对BCAT测验设计在实际应用中的选用提出了具体的意见与建议,以供实际应用者参考及借鉴.

Four new type of testing designs of computerized adaptive testing with bifactor model(BCAT) has been proposed on the basis of two traditional testing designs for BCAT.Two proposed optimality testing designs belong to the unidimensional BCAT,which are called as UBCAT_optimality1 and UBCAT_optimality2,respectively.Another two proposed optimality testing designs belongs to the multidimensional BCAT,which are called as MBCAT_optimality1 and MBCAT_optimality2,respectively.Results showed that:(i)The proposed four optimality designs for BCAT overall had higher parameter estimation precision of both general factor and special domain factor,than two exiting designs for BCAT.(ii)As for item bank exposure rate,the MBCAT designs were better than the UBCAT designs.The proposed MBCAT_optimality1 and the exiting MBCAT performed best in item exposure control.(iii)On test efficiency,the UBCAT designs used fewer items than those of the MBCAT designs.