中文摘要：

具有认知诊断功能的计算机化多阶段测验（CD-MST）是CDA和MST相结合的一种测验方式。由于CD-MST自适应频次较少, 初始阶段模块组建会影响整个测验的判准率。借鉴CD-CAT初始项目选取方法, 根据CDA和MST自身特点, 提出了7种CD-MST初始阶段模块组建方法, 分别是随机法、选题策略法、R＊矩阵法、CTTID法、CDI法、CTTIDR＊法和CDIR＊法。采用模拟研究对不同项目质量下7种方法的判准率进行了比较。研究结果表明, 当初始阶段结束时, 包含R＊矩阵的方法判准率显著高于其他方法, 尤其是CTTIDR＊法; 整个测验结束时, CTTIDR＊法较其他方法仍然有优势, CDIR＊法和R＊矩阵法结果较为接近。选题策略法在初始阶段结束时判准率较低, 甚至低于随机法, 整个测验结束时, 判准率同CDIR＊法和R＊矩阵法持平。4种项目质量对判准率影响较大, HD-HV题库下判准率最高, HD-LV次之, LD-HV较差, LD-LV最差。

英文摘要：

Cognitive Diagnostic Multistage Testing （CD-MST） is a new testing mode which combines MST with CDA. It has some advantages, such as improving instruction, allowing administers to check the pre-assembled test forms, allowing the examinee to review and revise answers. The initial stage module assemble methods is an important part for CD-MST. It is presented seven initial stage module assemble methods, they are Random method, Item Selection method, R＊ Matrix method, CTTID method, CDI method, CTTIDR＊ method and CDIR＊ method. 1） Random method means that select first stage items from item bank randomly, this method does not consider item parameters and attributes. It is easy and fast.2） Item selection method is first to randomly assign a knowledge state to examinee, and then select the initial stage items according to item selection indices （such as PWKL, SHE）. 3） R＊ matrix method is to select initial items from R-matrix. 4） CTTID method means that select high discriminatory items for the first stage. In DINA model, high discriminatory items are the items slipping and guessing parameters smaller.5） CDI method is presented by Henson and Douglas （2005）, it’s another item discrimination index, it considers item parameter and attribute.6） Item parameter and R matrix are the important elements which can influence the efficiency of testing, so here we consider discrimination and R matrix simultaneously to select first stage items, CTTIDR＊ method and 7） CDIR＊ method are from this consideration. A simulation research was used to verify these methods of CD-MST. CDM was DINA model, the number of examinee was 3000, item bank was 1240, the number of attributes was 5, stage number of CD-MST was 3. PCCR and ACCR were the evaluation indices. The quality of item bank was a contributory factor. The results showed that after the initial stage, under each of the quality of item banks, the PCCR of the methods with R＊ matrix were higher than others. When the tests were finished, CTTIDR＊ method wa