Abstract: Computerized adaptive test for cognitive diagnosis(CD_CAT) is the product of the theory and method of cognitive diagnosis(CD) and Computerized adaptive test(CAT), and is a new developing modern measurement field. The research on the selection strategy in CAT is always a focus topic internationally. While there are few reports about the research on the selection strategy in CD_CAT, and there are less reports about the research on initial items selection strategy in CD_CAT. Methods: Using Monte Carlo simulation, this paper compared five selection strategies and 2 initial items selection strategies with HO-DINA model. Conclusions: Different initial items selection methods and different selection strategies can affect the accurate of diagnosis and the precision of ability estimation. Totally, To two initial items selection strategies, the T-matrix method suggested in this paper is better than the traditional stochastic method; To five selection strategies, SL_GDI method is the best; and among the pairwise of initial items selection strategies and selection strategies, the pair of T-matrix method and SL_GDI is the best match.

Key words: cognitive diagnosis, Computerized adaptive test for cognitive diagnosis, HO-DINA model, item select strategy, initial items selection method

摘要： 计算机化认知诊断自适应测验(CD_CAT)是将认知诊断的基本理论、方法与计算机化自适应测验相结合的产物，是现代测量学发展的新领域。对于计算机化自适应测验（CAT）中的选题策略研究一直是国内外学者关注的问题，然而对于计算机化认知诊断自适应测验的选题策略研究却很少报导，而对于计算机化认知诊断自适应测验的初始题选取方法的研究却更少。本研究采用计算机模拟程序对HO-DINA模型下CD_CAT的五种选题策略及二种初始题选取方法进行研究。研究表明：不同初始题选取方法及选题策略均会影响对被试诊断的准确性及能力估计的精度；总体来看，对于二种初始题选取方法，本研究提出的“T阵法”优于传统的随机法；对于五种选题策略，SL_GDI法最优；初始题选取方法及选题策略的搭配中，“T阵法”和SL_GDI法的搭配最佳。

关键词: 认知诊断, 计算机化认知诊断自适应测验, HO-DINA模型, 选题策略, 初始题选取方法