CAJ | 학술논문

在计算机自适应测验中，对0—1评分模型按a－分层选题是高效安全的策略，但多级评分模型的项目难度／步骤参数有多个而无法直接应用这种选题策略。信息函数能够很好地综合项目所有参数及能力参数，但最大信息量选题策略会影响考试安全。本文提出一种变加权选题策略，它通过调用一个与信息量相关联的函数，该函数与信息量成正比，与区分度的某个幂函数成反比，从而达到既能综合项目所有参数又按a分层的效果。在GPCM模型下用蒙特卡罗实验进行比较研究，结果显示新的选题策略总体效果比已有相关结果好。
재계산궤자괄응측험중，대0—1평분모형안a－분층선제시고효안전적책략，단다급평분모형적항목난도／보취삼수유다개이무법직접응용저충선제책략。신식함수능구흔호지종합항목소유삼수급능력삼수，단최대신식량선제책략회영향고시안전。본문제출일충변가권선제책략，타통과조용일개여신식량상관련적함수，해함수여신식량성정비，여구분도적모개멱함수성반비，종이체도기능종합항목소유삼수우안a분층적효과。재GPCM모형하용몽특잡라실험진행비교연구，결과현시신적선제책략총체효과비이유상관결과호。
For dichotomous scoring, the a-stratified item selection strategy （a-STR） is an effective and safe method for computerized adaptive testing （CAT）. But it could not be applied to the polytomous scoring CAT, because there are too many parameters in the polytomous item response model to be comprehensive. Some domestic researchers employed the mean, the mode or one of the difficulties （or location parameters） to represent all of the difficulties or location parameters. In a word, the difficulty parameter in the dichotomous score item is replaced with some descriptive statistic of the difficulties or the location parameters then a-stratified item selection strategy is employed to polytomous CAT. It is well known that the Fisher information function is a good comprehension d all item parameters as well as the ability parameter. But the item selection strategy of maximum information index （MII） derogates the safe if MII is used at the early stage in the course of CAT for every examinee. Since the items with higher discrimination parameter may be frequently used and it causes the higher exposure rate for these items. A new item selection strategy named as the variable-weighted item selection strategy is proposed in this paper. Some functions of the Fisher information are employed to replace the a-STR. The new item selection strategy is comprehension of all information of the item parameters for polytomous items and plays the role of a-STR. Some comparisons of the new selection strategy with the other selection strategies combined with a-stratified based on generalized partial credit model （GPCM）. The results of Monte Carlo study show that the new selection strategy has the best effect.