计算机化自适应测验(CAT)是当前国内外考试改革的一个重要方向.该文对多维自适应测验中的能力估计和项目选择进行了一定的理论探讨,给出了贝叶斯估计中的项目选择算法.当A和x的范围已知时,这个算法把计算max|A+xx'|转换成计算max(x'A^-1x).
Computerized adaptive testing is an important direction of examination reform. The latent trait vector estimationand item selection in MAT are discussed in this paper. An algorithm of item selection for MAT is proposed. This algorithm changes computing max| A + xx'| into max(x'A^-1x) when A is known and the domain of x is known.